Add Neo4j-backed TypeScript analysis backend and formalize backend contracts

[rahlk]

Summary

Adds a Neo4j-backed analysis backend for TypeScript and formalizes the façade↔backend contract across all three languages so additional backends can be dropped in cleanly.

The latest codeanalyzer-typescript can emit its analysis graph to Neo4j (--emit neo4j, schema schema.neo4j.json). This PR adds an SDK backend that answers the exact same get_* query surface as the in-memory backend, but by running Cypher over that graph instead of walking the pydantic / NetworkX structures.

What's included

Neo4j TypeScript backend (cldk/analysis/typescript/neo4j/)

• TSNeo4jBackend — a drop-in alternative to TSCodeanalyzer implementing the full query surface (call graph, callers/callees, class hierarchy, call sites, decorators, symbol/method/field lookups, imports/exports/variables, …) via Cypher.
• It can populate the database for you over Bolt (runs the analyzer with --emit neo4j --neo4j-uri), or query a DB that is already loaded (build_db=False).
• Results are re-hydrated into the same cldk.models.typescript pydantic objects. Fields the projection inherently flattens (collapsed comments/type-params, aggregated import edges, the three round-tripped CALLS tag keys) are reconstructed best-effort and documented inline.
• Selected through the façade with a single optional parameter:

from cldk import CLDK
from cldk.analysis import AnalysisLevel
from cldk.analysis.typescript.neo4j import Neo4jConnectionConfig

analysis = CLDK(language="typescript").analysis(
    project_path="…",
    analysis_level=AnalysisLevel.call_graph,
    neo4j_config=Neo4jConnectionConfig(uri="bolt://localhost:7687", username="neo4j", password="…"),
)

No neo4j_config → the existing in-memory backend, unchanged.

Shared backend contracts (ABCs)

• TSAnalysisBackend, JavaAnalysisBackend, PythonAnalysisBackend — abstract bases declaring each façade's backend surface. The existing backends (TSCodeanalyzer, JCodeanalyzer, PyCodeanalyzer) and the new TSNeo4jBackend subclass them, so the relationship is enforced by the type system and at instantiation time rather than only by convention. This sets up equivalent Neo4j backends for Java/Python later.

Misc

• Optional neo4j extra (pip install cldk[neo4j]) for the driver.
• Regression test ensuring use_ray is forwarded through the Python façade (parity with the existing use_codeql / cache_dir guards).
• Tests: live-DB integration tests for the Neo4j backend (auto-skip when no Neo4j is reachable) mirroring the in-memory backend's sample-app expectations; no-DB backend-selection and contract tests for all three languages.

Notes

• The in-memory and Neo4j TypeScript backends were validated method-by-method against the same sample project for identical results.
• The Neo4j emission requires codeanalyzer-typescript 0.4.0+; the SDK pydantic models may need a follow-up bump to accept its newer entrypoints field for the in-memory path.

---

Generated by Claude Code