feat(graph): add C# language extractor

[doctor-ew]

Add code-graph extraction support for C# using tree-sitter-c-sharp (https://github.com/tree-sitter/tree-sitter-c-sharp), following the tree-sitter-only pattern (python.ts/rust.ts), not the TS compiler-based path.

• New src/graph/extraction/languages/csharp.ts with CSharpWalker extracting: classes/structs/interfaces/enums, methods (incl. constructors/destructors, static, async), properties, fields (const -> constant), parameters, namespaces (block + file-scoped, incl. nesting), using imports, attributes -> decorates, calls/instantiates/extends/implements.
• New fixture src/graph/__tests__/fixtures/sample.cs and 11 tests in extractor-csharp.test.ts.
• Register C# in languages/index.ts and grammars.ts (.cs extension).
• Vendor tree-sitter-c-sharp.wasm from tree-sitter-c-sharp@0.23.5 directly — not tree-sitter-wasms@0.1.12's copy of the same filename. That build is a different grammar revision: several fields node-types.json declares (variable_declarator.name, using_directive.name) resolve to undefined via childForFieldName against it. Verified by diffing behavior against both wasms; documented in docs/extractors.md.

Known, documented limitation: base_list has no field structure and doesn't syntactically distinguish a base class from an implemented interface. The extractor uses a first-listed-entry-is-the-base-class heuristic (extends for entry 0 on a class/struct, implements for the rest; every entry is extends on an interface) — a documented best-effort, not a semantic guarantee.

Not in scope for this pass: generics/type-parameter capture, LINQ query clauses, dedicated record/delegate node kinds (mapped to class/type_alias for now).

Validated against a real ~640-file C# codebase (a .NET Azure Functions calculation engine, external to this repo): 694 .cs-tree files, 0 partial/failed, 22,694 nodes / 25,250 edges, no regressions in the existing 533-test suite.

Fixes #155

---

What

Adds a C# (.cs) language extractor to the code graph.

Why

Closes #155 — C# had no extractor; .cs files were silently skipped by every graph build.

Type of change

• Bug fix
• New feature
• Refactor
• Docs
• CI/Tooling

How to test

1. npm test — runs the new extractor-csharp.test.ts (11 assertions) alongside the full suite (533/533 passing locally).
2. mex graph --root <a C# project> on a real .cs codebase and inspect .mex/graph.db.

Checklist

• Tests pass (npm test)
• No breaking changes (or documented below)
• Tested locally with a real project

Code-graph changes

• This PR targets main
• A linked issue agrees on the bounded extractor/resolver scope
• The change follows the frozen LanguageExtractor interface
• A focused fixture and assertions for the expected node/edge shape are included
• Any new grammar WASM, extension mapping, extractor, or resolver is registered
• No graph identity, reconciliation, schema, or drift-semantics changes are included

[Elmegaard]

Any update on getting this in?

[theDakshJaitly]

Any update on getting this in?

Yeah we'll review asap, should be out in next release

[theDakshJaitly]

Final review of df6a7d3: the declaration, identity, ownership, and reference-resolution findings are fixed with regression coverage. The complete PR was checked against current main; graph tests, typechecks, builds, evaluators, and CI pass. The remaining partial C# support limits are documented. No remaining blocking findings.