Complete TypeScript human ground-truth audit

[DavidBakerEffendi]

Description

Completes the first human ground-truth review of all 12 TypeScript benchmark cases as part of #58. The reviewed contracts now state explicit navigation operations, conservative static-analysis rationale, and post-review TypeScript Language Server and Bifrost outcomes.

Key Changes:

• Classify concrete calls, constructions, type references, JSX references, contextual properties, barrel aliases, and shadowing exclusions across the TypeScript baseline, precision, and LSP-parity corpora.
• Allow reviewed navigation alternates with different symbol kinds, so a canonical class target can coexist with an explicit constructor target without weakening the required destination.
• Migrate the previously reviewed JavaScript class-construction control to the cross-kind alternate model and update the rendered case comparison.

Touch Points:

• benchmarks/cases/typescript-baseline.yaml
• benchmarks/cases/typescript-precision.yaml
• benchmarks/cases/typescript-lsp-parity.yaml
• benchmarks/cases/javascript-baseline.yaml
• benchmarks/reviews/2026-07-17-DavidBakerEffendi.md
• benchmarks/README.md
• src/lib.rs
• docs/src/content/docs/results/case-comparison.md

Validation:

• cargo fmt --check
• cargo test (106 tests passed)
• cargo run -- validate benchmarks/cases (35 documents)
• TypeScript 5.9.3 compilation for all three TypeScript fixture roots
• Isolated full-document TypeScript Language Server 5.3.0 and Bifrost 0.8.9 comparisons
• npm run check
• npm run build
• Rendered inspection of /results/case-comparison/