Pre-fix `vera/checker/registration.py::_register_alias` resolved
aliases one at a time; when `type A = B` was processed before
`B` was registered, the forward-reference fallback in
`_resolve_type` returned a placeholder rather than chasing the
chain.  The resolved-type representation reached the
post-registration state with no observable cycle.

Codegen then stored the raw AST `type_expr` and
`vera/codegen/core.py::_type_expr_to_wasm_type` chased the chain
through the AST, blowing the stack with `RecursionError:
maximum recursion depth exceeded` — pointing at codegen
internals rather than the actual program bug.

Post-fix `_register_all` calls a new `_check_alias_cycles` pass
that walks every alias's AST `type_expr` chain (following
`NamedType`-of-alias references through `RefinementType`
wrappers, mirroring codegen's recursion shape) and emits
`[E132]` ("Cyclic type alias") with:
- the originating decl location
- the full cycle path (`A -> B -> C -> A`)
- a `Fix:` paragraph pointing at `data`-declared ADTs as the
  alternative for self-referential types
- spec ref to §4.3 "Type Aliases"

The walker mirrors `_type_expr_to_wasm_type`'s recursion shape
deliberately — it follows the same chain that would otherwise
crash codegen, so a cycle the checker detects is exactly a
cycle that would have crashed.  Cycles passing *through* an
`Array<T>` / `FnType` / tuple terminate before reaching another
alias (because codegen returns a concrete WASM type at those
nodes without recursing into children), so they aren't the ones
this fix targets — and they don't crash codegen either.

Defensive cycle guards on the alias-walking helpers in
`vera/wasm/inference.py` (closed in #633) remain as
belt-and-braces for the same family.

Tests: 4 new tests in `tests/test_checker.py::TestResolutionCoverage`:
- `test_cyclic_alias_two_way_e132` (canonical case)
- `test_cyclic_alias_self_e132` (`type A = A`)
- `test_cyclic_alias_three_way_e132` (`A -> B -> C -> A`)
- `test_acyclic_alias_chain_ok` (no false-positive on
  `type IntAlias = Int; type Pair = IntAlias`)

Validation
- pytest tests/ -q → 3,810 passed, 14 skipped
- mypy vera/ → clean

Closes #648

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

CR-1 (`tests/test_checker.py:5508-5575`) — add
`test_cyclic_alias_refinement_e132` covering the refinement-
wrapper cycle `type A = { @b | true }; type B = A`.  Pins the
`_alias_chain_target` helper's `RefinementType.base_type`
peeling — codegen's `_type_expr_to_wasm_type` recurses through
refinements unconditionally, so a cycle hidden behind one is
still a codegen-crash cycle and must be intercepted by the
checker.  Verified the case fires `[E132]` correctly against
the current walker; this test pins the contract.

CI lint (`scripts/check_version_sync.py`) — `README.md`'s
`v0.0.148`/`148 releases`/`3,820 tests` row was missed in the
version bump.  Updated to `v0.0.149` / `149 releases` / `3,825
tests` (the latter accounts for both the four #648 tests
landed in the parent commit and the new refinement-cycle test
in this commit).  TESTING.md and ROADMAP.md counts updated
accordingly.

Validation
- pytest tests/test_checker.py -k "cyclic_alias or acyclic_alias" -v
  → 5 passed (was 4)
- python scripts/check_version_sync.py → consistent across 6 files
- python scripts/check_doc_counts.py → 3,825 tests / 29 files (live)

Refs #648 #665

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

CR-2 (`tests/test_checker.py:5508-5585`) — pin diagnostic
payload on a representative test.

The four cyclic-alias tests landed in `f23e261` asserted only
that an `[E132]` diagnostic existed — not its contents.  A
future refactor that re-routes through a sibling code with
matching error_code but different rendering would silently
regress the user-facing message.

Fix: on `test_cyclic_alias_two_way_e132` (the canonical case),
add two payload assertions:

1. `"A -> B -> A" in e132[0].description` — pins that the cycle
   path is rendered so the user can identify *which* aliases
   participate.

2. `"data" in e132[0].fix` — pins the `data`-declared ADT
   suggestion in the fix hint, since `data` is the canonical
   alternative when self-reference is what the user wanted.

The other three E132 tests still check error_code only — the
payload contract these assertions embody is uniform across all
of them, so duplicating it on every test would be cargo-cult.
Documented this scope decision in the canonical test's
docstring so a future reader can find the payload assertions
without grepping all four.

Note on field name: CR's comment referenced `fix_hint`; the
actual `Diagnostic` attribute is `fix` (see `vera/errors.py:58`).
Intent of the suggestion preserved; field name corrected.

Validation
- pytest tests/test_checker.py::TestResolutionCoverage::test_cyclic_alias_two_way_e132 -v
  → 1 passed

Refs #648 #665

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