[tests] parser diff harness — parser-ts vs parser-native AST equality on fixture corpus

[cs01]

Goal

Catch AST divergence between `src/parser-ts/` (TS compiler API, node compiler) and `src/parser-native/` (tree-sitter, self-hosted). Standalone regression gate. Pre-req for eventually retiring parser-ts.

Scope (~300-400 LOC)

• `tests/parser-diff.test.ts` (~150) — iterate `tests/fixtures/**/*.ts`, parse with both, deep-equal normalized JSON
• `src/parser-diff/normalize.ts` (~100) — strip `loc`/`range`, canonicalize field order, drop parser-specific metadata
• `src/parser-diff/allowlist.ts` (~50) — known divergences w/ reason + linked issue

Behavior

• Fail on new divergence
• Allowlisted entries logged, not failed
• `npm run test:parser-diff` standalone target

Expected outcome

First run likely surfaces 5-20 real divergences. Each becomes either a parser-native fix or allowlist entry linking a new bug issue.

Non-goals

• Translation layer between parsers
• Rewriting either parser
• Codegen changes

Risk

Low — pure test infrastructure.

Related

• [meta] refactor plan tracker: 5-step rearchitecture (resolver memo → IRBuilder → TypedAST annotator) #640 (rearchitecture tracker, orthogonal)
• parser: rejects arr[i](args) — ElementAccessExpression as callee #589 (historical parser divergence, closed)

[cs01]

closing stale/premature issue — will re-file when actively working on this