PR boxlite-ai#678 shipped e2e pytest cases under scripts/test/e2e/cases/, a
fresh location that doesn't match the repo's existing per-SDK test
layout (src/cli/tests/*.rs, sdks/python/tests/, sdks/node/tests/,
sdks/c/tests/, sdks/go/*_test.go).

This PR moves the cases to match. Driver scripts stay in
scripts/test/e2e/ (cross-SDK infra); test files now live next to
the SDK they exercise:

  scripts/test/e2e/cases/test_c_entry.py    → sdks/c/tests/e2e/
  scripts/test/e2e/cases/test_go_entry.py   → sdks/go/tests/e2e/
  scripts/test/e2e/cases/test_node_entry.py → sdks/node/tests/e2e/
  scripts/test/e2e/cases/test_cli_entry.py  → src/cli/tests/e2e/
  scripts/test/e2e/cases/test_*.py (Python pytest)
                                            → sdks/python/tests/e2e/
  scripts/test/e2e/cases/conftest.py        → sdks/python/tests/e2e/
  scripts/test/e2e/sdks/{c,go,node}/e2e_basic.{c,go,ts}
                                            → sdks/{c,go,node}/tests/e2e/

Stays in scripts/test/e2e/: bootstrap.sh, teardown.sh, run.sh,
two_sided.sh, fixture_setup.py, pytest.ini, README.md.

Supporting wiring (no test-logic change):
  - run.sh: pytest invoked with the 5 SDK e2e dirs as explicit args
    (testpaths in pytest.ini can't resolve outside its rootdir)
  - pytest.ini: comment explaining the testpaths omission
  - .github/workflows/e2e-stack.yml: paths filter expanded to all
    new e2e locations
  - README.md: layout map updated

Pure rename + supporting wiring. Zero new test logic, zero source
changes. Originally part of boxlite-ai#681 which also bundled 21 new test
cases + 2 fixes; this split-out covers reorganize only — the rest
ships in follow-up PRs.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

21 new e2e cases under sdks/python/tests/e2e/ that pin contracts the
existing local-FFI integration tests can't reach. All ship with
`xfail(strict=True)` for the 10 known production bugs they pin:

  test_error_code_mapping.py  10 cases  4 xfails
  test_quota_enforcement.py    5 cases  5 xfails (module-level)
  test_runner_concurrency.py   6 cases  1 xfail
                              --       --
                              21       10

Each xfail's reason= string carries a `file:line` pointer to the bug
location and a short root-cause note, so a follow-up fix-PR can grep
for the right reason, drop the marker, and watch the test go green.
`strict=True` means an unexpected xpass becomes a CI failure — i.e.
the moment any of these bugs is silently fixed we'll notice.

xfail coverage map (which PR removes which marker):

  test_exec_after_box_removed_is_typed_error
    → PR C (runner exec error mapping: IsNotFound → 404)

  test_invalid_argument_zero_cpu_returns_400          )
  test_invalid_argument_negative_memory_returns_400   ) needs API
  test_oversized_cpu_returns_400                      ) ValidationPipe
  5x test_quota_enforcement.py                        ) + per-sandbox
                                                       ) quota fix
                                                       ) (separate
                                                       ) follow-up)

  test_execution_invalid_command_returns_422
    → needs Rust spawn-failed → ErrExecution mapping fix
      (separate follow-up; not in C or D)

Split out of boxlite-ai#681 — this is **PR B of 4**:
  A (boxlite-ai#682) reorganize     — merged before this rebases cleanly
  B (this)  21 new cases
  C         runner exec error mapping fix
  D         FFI exec drain race fix

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

`POST /v1/{prefix}/boxes/{id}/exec` was leaking HTTP 500 when the
underlying box was already gone / stopped / in a non-runnable state
by the time the runner tried to spawn. Two related observable bugs:

  exec on a removed box   → 500 "spawn_failed: build failed"
  exec on a stopped box   → 500 (same shape)
  exec on box mid-rebuild → 500 (same shape)

Root cause: `BoxliteExec` in `apps/runner/pkg/api/controllers/boxlite_exec.go`
called `execManager.Start(...)` and unconditionally wrapped any non-nil
error as 500. The SDK Start path already tunnels typed errors from the
Rust core (sdks/go/errors.go: IsNotFound / IsStopped / IsInvalidState)
when the box state changed under the request — they just weren't being
read.

Fix: extend `classifyExecError` to peek for those SDK typed errors and
return the canonical mapping from `src/shared/src/errors.rs:198-280`:

  IsNotFound       → 404 Not Found
  IsStopped        → 409 Conflict
  IsInvalidState   → 409 Conflict
  (other / no SDK match) → 500 (unchanged)

Also routes the `Start()` error site through `classifyExecError`
instead of hard-coding 500.

Flips the xfail on `test_exec_after_box_removed_is_typed_error`
(sdks/python/tests/e2e/test_runner_concurrency.py) — that case now
xpasses, which under strict=True trips the marker and fails CI; the
fix here removes the marker so it lands green.

Other related xfails in test_error_code_mapping.py are NOT addressed
by this PR (they need DTO validation / quota / Rust spawn-failed
reclassification — separate follow-ups).

Split out of boxlite-ai#681 — this is **PR C of 4**:
  A (boxlite-ai#682) reorganize
  B (boxlite-ai#683) 21 new e2e cases (xfails pin bugs)
  C (this) runner exec error mapping fix + flip 1 xfail
  D        FFI exec drain race fix

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

The Go SDK's `box.Exec` would occasionally return with stdout chunks
still in flight — the user's callback got `OnExit` (or the wait gRPC
reply) before the matching `OnStdout`/`OnStderr` chunks landed on
the queue. From the caller's perspective, the exec had finished but
the last few bytes of stdout were missing.

Root cause: `execution_wait` and `exit_pump` pushed their terminal
events as soon as the underlying process exited, racing the still-
draining stream pumps. The FFI queue is ordered, but the wait/exit
events were entering it ahead of late stdout chunks.

Fix: track a `streams_pending` count on `ExecutionHandle` instead of
the previous per-pump receiver list. Both `exit_pump` and the
`execution_wait` task now await `streams_pending → 0` (via a
`tokio::sync::Notify` notification + standard register-then-check
pattern to avoid post-notification miss) before pushing their
terminal events.

Deterministic ordering tests added inline in execution.rs cover:

  - stdout-only exec: all chunks land before Exit
  - mixed stdout/stderr: every chunk lands before Wait
  - process exit before pump completion: terminal event waits
  - pump completion before process exit: terminal event fires
    immediately (no spurious wait)

Side effect on the existing e2e suite: `test_p0_6_exec_stdout_race`
in sdks/python/tests/e2e/ (the regression test for boxlite-ai#563) flips from
~90% loss against stock 0.9.5 to 0% loss against this PR. That test
is NOT in the PR-B xfail set (it was pre-existing), so no markers
need flipping.

Other PR-B xfails (DTO validation, quota, spawn-failed mapping) are
NOT addressed here — separate root causes, separate follow-ups.

Split out of boxlite-ai#681 — this is **PR D of 4**:
  A (boxlite-ai#682) reorganize
  B (boxlite-ai#683) 21 new e2e cases (xfails pin bugs)
  C (boxlite-ai#684) runner exec error mapping fix + flip 1 xfail
  D (this) FFI exec drain race fix

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

PR boxlite-ai#678 shipped e2e pytest cases under scripts/test/e2e/cases/, a
fresh location that doesn't match the repo's existing per-SDK test
layout (src/cli/tests/*.rs, sdks/python/tests/, sdks/node/tests/,
sdks/c/tests/, sdks/go/*_test.go).

This PR moves the cases to match. Driver scripts stay in
scripts/test/e2e/ (cross-SDK infra); test files now live next to
the SDK they exercise:

  scripts/test/e2e/cases/test_c_entry.py    → sdks/c/tests/e2e/
  scripts/test/e2e/cases/test_go_entry.py   → sdks/go/tests/e2e/
  scripts/test/e2e/cases/test_node_entry.py → sdks/node/tests/e2e/
  scripts/test/e2e/cases/test_cli_entry.py  → src/cli/tests/e2e/
  scripts/test/e2e/cases/test_*.py (Python pytest)
                                            → sdks/python/tests/e2e/
  scripts/test/e2e/cases/conftest.py        → sdks/python/tests/e2e/
  scripts/test/e2e/sdks/{c,go,node}/e2e_basic.{c,go,ts}
                                            → sdks/{c,go,node}/tests/e2e/

Stays in scripts/test/e2e/: bootstrap.sh, teardown.sh, run.sh,
two_sided.sh, fixture_setup.py, pytest.ini, README.md.

Supporting wiring (forced by the renames):
  - run.sh: pytest invoked with the 5 SDK e2e dirs as explicit args
    (testpaths in pytest.ini can't resolve outside its rootdir)
  - pytest.ini: comment explaining the testpaths omission
  - .github/workflows/e2e-stack.yml: paths filter expanded to all
    new e2e locations
  - README.md: layout map updated

Two collisions the rename exposed (both fixed in this PR):

  1. SDK unit-test pytest workflow auto-discovers sdks/python/tests/
     and would import the new e2e/ subdir, whose conftest needs
     Python 3.11+ tomllib and a live API stack the unit-test
     workflow doesn't provide. Fix: norecursedirs = e2e in
     sdks/python/pytest.ini. The e2e runner passes the dir
     explicitly on the CLI which bypasses norecursedirs.

  2. pytest's default --import-mode=prepend puts each test file's
     rootpath on sys.path. For sdks/python/tests/e2e/test_*.py the
     rootpath becomes sdks/python/, which then shadows the installed
     boxlite wheel with the source-tree stub package
     (sdks/python/boxlite/__init__.py — missing the compiled .so so
     Rust-bound symbols like BoxliteRestOptions are absent). Every
     fixture using boxlite.BoxliteRestOptions failed at setup with
     AttributeError. Fix: --import-mode=importlib in
     scripts/test/e2e/pytest.ini (modern mode, no sys.path
     injection). Tests that previously did `from conftest import
     drain` couldn't keep doing that under importlib mode (conftest
     no longer on sys.path) — extracted drain/collect_stream/
     stdout_line_count from conftest into
     scripts/test/e2e/lib/e2e_helpers.py (alongside path_verification
     which already used the same lib/ + sys.path.insert pattern),
     and updated the 8 test files' imports.

Verified locally on Ubuntu 22.04 + Python 3.11 + maturin-built
boxlite wheel — make test:e2e against PR-A: 21 passed, 7 failed
(the 7 are real production bugs the suite is designed to catch:
PR-C fixes runner exec error mapping, PR-D fixes FFI drain race).
Before this PR's collision fix: 0 passed, 32 fixture-setup errors.

Originally part of boxlite-ai#681 which also bundled 21 new test cases + 2
fixes; this split-out covers reorganize only — the rest ships in
follow-up PRs.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

21 new e2e cases under sdks/python/tests/e2e/ that pin contracts the
existing local-FFI integration tests can't reach. All ship with
`xfail(strict=True)` for the 10 known production bugs they pin:

  test_error_code_mapping.py  10 cases  4 xfails
  test_quota_enforcement.py    5 cases  5 xfails (module-level)
  test_runner_concurrency.py   6 cases  1 xfail
                              --       --
                              21       10

Each xfail's reason= string carries a `file:line` pointer to the bug
location and a short root-cause note, so a follow-up fix-PR can grep
for the right reason, drop the marker, and watch the test go green.
`strict=True` means an unexpected xpass becomes a CI failure — i.e.
the moment any of these bugs is silently fixed we'll notice.

xfail coverage map (which PR removes which marker):

  test_exec_after_box_removed_is_typed_error
    → PR C (runner exec error mapping: IsNotFound → 404)

  test_invalid_argument_zero_cpu_returns_400          )
  test_invalid_argument_negative_memory_returns_400   ) needs API
  test_oversized_cpu_returns_400                      ) ValidationPipe
  5x test_quota_enforcement.py                        ) + per-sandbox
                                                       ) quota fix
                                                       ) (separate
                                                       ) follow-up)

  test_execution_invalid_command_returns_422
    → needs Rust spawn-failed → ErrExecution mapping fix
      (separate follow-up; not in C or D)

Split out of boxlite-ai#681 — this is **PR B of 4**:
  A (boxlite-ai#682) reorganize     — merged before this rebases cleanly
  B (this)  21 new cases
  C         runner exec error mapping fix
  D         FFI exec drain race fix

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

`POST /v1/{prefix}/boxes/{id}/exec` was leaking HTTP 500 when the
underlying box was already gone / stopped / in a non-runnable state
by the time the runner tried to spawn. Two related observable bugs:

  exec on a removed box   → 500 "spawn_failed: build failed"
  exec on a stopped box   → 500 (same shape)
  exec on box mid-rebuild → 500 (same shape)

Root cause: `BoxliteExec` in `apps/runner/pkg/api/controllers/boxlite_exec.go`
called `execManager.Start(...)` and unconditionally wrapped any non-nil
error as 500. The SDK Start path already tunnels typed errors from the
Rust core (sdks/go/errors.go: IsNotFound / IsStopped / IsInvalidState)
when the box state changed under the request — they just weren't being
read.

Fix: extend `classifyExecError` to peek for those SDK typed errors and
return the canonical mapping from `src/shared/src/errors.rs:198-280`:

  IsNotFound       → 404 Not Found
  IsStopped        → 409 Conflict
  IsInvalidState   → 409 Conflict
  (other / no SDK match) → 500 (unchanged)

Also routes the `Start()` error site through `classifyExecError`
instead of hard-coding 500.

Flips the xfail on `test_exec_after_box_removed_is_typed_error`
(sdks/python/tests/e2e/test_runner_concurrency.py) — that case now
xpasses, which under strict=True trips the marker and fails CI; the
fix here removes the marker so it lands green.

Other related xfails in test_error_code_mapping.py are NOT addressed
by this PR (they need DTO validation / quota / Rust spawn-failed
reclassification — separate follow-ups).

Split out of boxlite-ai#681 — this is **PR C of 4**:
  A (boxlite-ai#682) reorganize
  B (boxlite-ai#683) 21 new e2e cases (xfails pin bugs)
  C (this) runner exec error mapping fix + flip 1 xfail
  D        FFI exec drain race fix

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

The Go SDK's `box.Exec` would occasionally return with stdout chunks
still in flight — the user's callback got `OnExit` (or the wait gRPC
reply) before the matching `OnStdout`/`OnStderr` chunks landed on
the queue. From the caller's perspective, the exec had finished but
the last few bytes of stdout were missing.

Root cause: `execution_wait` and `exit_pump` pushed their terminal
events as soon as the underlying process exited, racing the still-
draining stream pumps. The FFI queue is ordered, but the wait/exit
events were entering it ahead of late stdout chunks.

Fix: track a `streams_pending` count on `ExecutionHandle` instead of
the previous per-pump receiver list. Both `exit_pump` and the
`execution_wait` task now await `streams_pending → 0` (via a
`tokio::sync::Notify` notification + standard register-then-check
pattern to avoid post-notification miss) before pushing their
terminal events.

Deterministic ordering tests added inline in execution.rs cover:

  - stdout-only exec: all chunks land before Exit
  - mixed stdout/stderr: every chunk lands before Wait
  - process exit before pump completion: terminal event waits
  - pump completion before process exit: terminal event fires
    immediately (no spurious wait)

Side effect on the existing e2e suite: `test_p0_6_exec_stdout_race`
in sdks/python/tests/e2e/ (the regression test for boxlite-ai#563) flips from
~90% loss against stock 0.9.5 to 0% loss against this PR. That test
is NOT in the PR-B xfail set (it was pre-existing), so no markers
need flipping.

Other PR-B xfails (DTO validation, quota, spawn-failed mapping) are
NOT addressed here — separate root causes, separate follow-ups.

Split out of boxlite-ai#681 — this is **PR D of 4**:
  A (boxlite-ai#682) reorganize
  B (boxlite-ai#683) 21 new e2e cases (xfails pin bugs)
  C (boxlite-ai#684) runner exec error mapping fix + flip 1 xfail
  D (this) FFI exec drain race fix

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>