TypeScript parity: port the four bulk/projected accessors (callables overview, method bodies, decorated callables, callsites) from Python

[rahlk]

The TypeScript surface has none of the four bulk/projected accessors Python shipped in #180/#181, on either line (verified at v1.4.3 and release/2.0):

Python (template)	TypeScript
get_callables_overview() -> List[PyCallableOverview] (backend.py:152, facade python_analysis.py:526)	missing
get_method_bodies(signatures) -> Dict[str, str] (backend.py:157)	missing
get_decorated_callables(markers) -> List[PyCallableOverview] (backend.py:162)	missing
get_callsites_for(signatures) -> Dict[str, List[PyCallsite]] (backend.py:167)	missing
PyCallableOverview (cldk/models/python/projections.py:32)	no TS projection model

Scope (all four, one change)

• TSCallableOverview projection model — the one real design decision: the field list. Open questions to settle before implementation (design-gate item, light pass since Bulk/projection accessors for the Python facade to avoid N+1 reconstruction on the Neo4j backend #180/feat(python): bulk/projected accessors for the Python facade (#180) #181 locked the shape): how TS-specific facets project (arrow/anonymous callable kinds, owning class vs interface, export visibility, decorators), and — same division of labor as Python — the overview excludes body/code text (get_method_bodies owns that).
• Four @abstractmethods on TSAnalysisBackend, implemented on BOTH backends: TSCodeanalyzer (in-process over its existing _index() lookups) and TSNeo4jBackend (set-at-a-time, one Cypher round trip each — that is the point of the bulk surface).
• Four thin facade delegates on TypeScriptAnalysis.
• Dual-backend parity tests per accessor (template: the existing Python parity suite).

Version routing

Additive public API ⇒ minor bump: lands as 1.5.0 on the 1.x line (needs a maintenance branch cut from v1.4.3 — none exists yet), and must ALSO land on release/2.0 or 2.0.0 ships a parity regression against 1.5.

Follows the #180/#181 design; sibling of the Python bulk-accessor surface.

[rahlk]

Design locked (2026-07-27): (1) overview carries NO code field — bodies stay behind get_method_bodies; (2) facets = parity core + is_exported/is_async/is_static/accessibility, kind = analyzer-native 7-value vocabulary passed through; (3) owner = owner_signature + owner_kind ('class'|'interface') pair, None/None for module-level and nested; (4) 1.x line first: release/1.4 cut from v1.4.3, ships as 1.5.0, then forward-port to release/2.0. Line-dependent constraint found during design review: the 1.4-era cants emit carries :Decorator/DECORATED_BY (SDK already reconstructs), but the 2.0-era emit projects no decorators — the forward-port needs an upstream cants issue (persist decorators: string[] on TSCallable) and its Neo4j leg degrades honestly until that lands. Implementation starting on feat/issue-298-ts-bulk-accessors.

[rahlk]

Spec amendment (final-review ruling, 2026-07-27): get_method_bodies' omission rule is mechanical — omit iff the callable's code is None (implicit constructors, synthesized callables), plus unknown signatures. The design doc's earlier 'interface stubs are omitted' wording rested on a false premise: the cants 0.4.3 emit DOES carry source text for interface method signatures, so stubs with code ARE returned. Both backends implement exactly this; every returned value is a real str. The 2.0 forward-port must implement THIS contract, not the original spec wording. Also filed during final review: #300 (latent getter/setter signature-collision parity break, behavior fix is follow-up).