Closes #660.  Investigates and documents #661 (not reachable today).

## #660 — Type-checker accepts wrong-arity parameterised aliases

Pre-fix `vera/checker/resolution.py::_resolve_type` silently
truncated `zip(alias.type_params, te.type_args)` on length
mismatch, leaving alias-local type-vars unsubstituted.
Downstream codegen leaked literal alias-local names into mono
suffixes (`option_map$Int_B` instead of `option_map$Int_Int`)
and the call site referenced a non-existent function-table
entry → `unknown table 0: table index out of bounds` at
runtime.  Surfaced by the #659 multi-agent review (finding F3).

### Fix

In `_resolve_type`'s type-alias branch, before substitution,
check `len(te.type_args) == len(alias.type_params)`.  On
mismatch, emit `[E133]` ("Type alias arity mismatch") with:
- The alias name + expected/supplied counts
- A rationale paragraph explaining the codegen consequence
- A `Fix:` paragraph suggesting the missing or extra type
  arguments (with the alias's declared param names quoted)
- Spec reference: Chapter 2, Section 2.4 "Type Aliases"

Returns `UnknownType()` so downstream checking continues
without cascading errors.

### Defensive comments removed

The two defensive comments left in `vera/codegen/monomorphize.py::_resolve_arg_fn_shape`
and `vera/wasm/calls.py::_resolve_arg_fn_shape_wasm` in PR #659
(documenting the latent gap) are now trimmed to one-line
"arity enforced upstream by checker" cross-references.  Their
verbose explanation is no longer needed since the gap is closed.

### New tests

`tests/test_checker.py::TestResolutionCoverage` gains three
tests:

1. `test_alias_arity_mismatch_too_few_e133` — `@Pair<Int>`
   against `Pair<A, B>` rejected with E133
2. `test_alias_arity_mismatch_too_many_e133` — `@Single<Int, Bool>`
   against `Single<T>` rejected with E133 (symmetric case)
3. `test_alias_zero_args_when_zero_expected_ok` — non-parameterised
   alias accepts no type-args (pins the no-false-positive case)

## #661 — investigated; not reachable today; invariant pinned

The agent finding flagged a hypothetical scenario: if two
modules both declare `forall<T> fn shared_name(...)`, the
template-warning suppression set keys on the bare base name and
could over-suppress.  Investigation found this isn't reachable
today because:

1. Pass 2.5 in `compile_program` (lines 519-530) explicitly
   skips imported FnDecls whose names are already in
   `fn_visibility` (= local declarations).  An imported forall
   decl with the same name as a local one is dropped before
   its template warning could be emitted.
2. `forall_decl_names` is built from `program.declarations`
   only, never from imports.  Only local forall decls are
   eligible for suppression.

At most one template warning per base name lands in
`self.diagnostics`, so bare-name matching cannot cross-suppress.

### Added an explanatory comment + invariant-pinning tests

The `compile_program` suppression block now has a comment
explicitly naming the two preconditions that keep bare-name
keying safe, and the trigger conditions that would invalidate
the invariant (loosened Pass 2.5 dedup, or mono pipeline
starting to carry module attribution).

`tests/test_codegen_modules.py::TestCrossModuleNameCollision661`
adds two tests:

1. `test_cross_module_forall_name_shadow_compiles_and_runs` —
   the name-shadow fixture compiles and runs (returns 42).
2. `test_suppression_does_not_cross_modules` — no [E602]
   warning for `shared_name` despite the cross-module shadow.

If Pass 2.5's dedup or `forall_decl_names`' scope ever changes,
these tests flag it.

## Release prep (v0.0.148)

Bundled per the ABSOLUTE RULE:
- `vera/__init__.py` + `pyproject.toml` + `uv.lock`: 0.0.147 → 0.0.148
- `CHANGELOG.md`: `[0.0.148] - 2026-05-12` with `### Fixed` (#660)
  and `### Internal` (#661) entries + link refs
- `HISTORY.md` Stage 12 row: v0.0.148 + roll-up counter 147 → 148
- `docs/index.html`: version tag
- `README.md` / `TESTING.md` / `ROADMAP.md`: count refreshes
  (3,820 tests, 148 releases)
- `vera/errors.py`: new `E133` entry between `E131` and `E140`

## Verification

- mypy clean (59 source files)
- pytest: 3,806 passed, 14 skipped (was 3,801 + 5 new tests)
- All 86 conformance programs pass
- All 34 examples pass `vera check` + `vera verify`
- e602 gate: 116 files clean, 5 allowlist matched, 0 stale
- version-sync, doc-counts: consistent

Refs #604 #659 #660 #661

Co-Authored-By: Claude <noreply@anthropic.invalid>

PR #663 (v0.0.147) closed #628 but the matching row in
KNOWN_ISSUES.md's Bugs table was missed in that PR's release
prep.  Removed now while #664 is still open.

Other surfaces verified clean for #628 / #660 / #661:
- `ROADMAP.md`: no burn-down rows for these (already swept)
- `vera/README.md`: no mentions
- `spec/*.md`: no mentions
- `examples/*.vera`: no workaround code
- `tests/conformance/*.vera`: no workaround code
- `tests/test_codegen_modules.py` / `tests/test_checker.py`:
  references are the new regression tests for #628 / #660 / #661
  — intentional and kept

Limitation count: 34 → 33.

Refs #628 #660 #661 #664

Co-Authored-By: Claude <noreply@anthropic.invalid>

CR-3 (`tests/test_codegen_modules.py:592`) — guard against
silent pass-on-failure.

`test_suppression_does_not_cross_modules` inspected only the
filtered `shared_warnings` list.  If `compile()` errored out, the
warning list would be empty and the `assert not shared_warnings`
would silently pass — defeating the test's intent (which is to
verify the suppression filter doesn't drop a legitimate cross-
module warning).

Fix: assert `result.ok` immediately after the `compile(...)` call,
before computing the warning subsets.  Mirrors the explicit error
check already present in the sibling
`test_cross_module_forall_name_shadow_compiles_and_runs` test
(lines 536-540).

Validation
- pytest tests/test_codegen_modules.py::TestCrossModuleNameCollision661 -v
  → 2 passed

Refs #661 #664

Co-Authored-By: Claude <noreply@anthropic.invalid>