feat: exclude globs and extra PHP extensions for indexing

[syntlyx]

Each commit is one logical change and carries its own description:

1. fix: classify vendor files correctly under a symlinked workspace root
2. feat: exclude globs and extra PHP extensions for indexing
3. fix: index Drupal test directories so test base classes resolve

Summary

Ships the server side of backlog item X9 (docs/todo/indexing.md): user-configurable file filters for workspace discovery.

[indexing]
exclude = ["generated", "web/sites/default/files", "tests/"]
extensions = ["module", "install", "theme"]

• exclude — patterns in gitignore syntax relative to the workspace root: a bare name matches at any depth, a pattern containing / anchors to the root, a trailing / restricts to directories, a leading ! re-includes. Honored by every workspace walker (fallback full scan, PSR-4/vendor scans, the Drupal web-root scan, preload, go-to-implementation) and by the file watcher, so events for excluded paths are dropped too. Excludes apply to background discovery only — files opened in the editor are always served.
• extensions — extra file extensions (without the dot) treated as PHP source, e.g. Drupal's .module/.inc/.theme outside the auto-detected web root. Each gets its own didChangeWatchedFiles watcher so edits refresh the index.

Implementation notes

• Patterns compile once into a new classmap_scanner::IndexFilters (backed by ignore::gitignore::GitignoreBuilder), cached on the Backend, invalidated on config reload — no glob compilation on any per-file path. Empty config is a single Option branch in the walkers, so the default configuration costs nothing.
• GitignoreBuilder was chosen over ignore's Override: Override's whitelist-first semantics would invert the meaning of an exclude list containing ! re-includes.
• Invalid patterns are skipped with a warning (same policy as [[diagnostics.ignore]]), never failing config load.
• config-schema.json and docs/configuration.md updated; changelog entry added.
• X9 in the backlog is rewritten to its remaining scope: forwarding the editor's files.exclude / files.associations from the editor extensions to the server (initializationOptions / workspace/configuration), which needs client-side work.

Bundled fixes (separate commits)

Drupal test base classes resolve. The Drupal web-root scanner skipped every directory named tests/Tests, but module tests extend base classes that live under core/tests/ (Drupal\Tests\UnitTestCase, Drupal\KernelTests\KernelTestBase, …), and the gitignore-aware workspace scan never reaches core/ either — so on a real Drupal project every module test reported Class 'Drupal\Tests\UnitTestCase' not found (very visible now that workspace-wide diagnostics are on by default). Test directories are indexed by default; the classmap's first-wins merge handles duplicate fixture FQNs, and [indexing] exclude replaces the hardcoded name filter for projects that want a smaller index.

Vendor classification under a symlinked workspace root. Pre-existing bug reproducible on clean main: on a workspace whose real path goes through a symlink (macOS /var → /private/var, link-farm checkouts), vendor files were classified as project code — hover lost the package provenance badge, vendor providers were scanned as user code, and the analyse CLI walked the entire vendor tree. Scanned paths are canonicalized but the vendor dir was stored raw-only; add_vendor_dir now records the canonical form alongside the raw one (mirroring what the vendor URI prefixes already do), and analyse's skip list is canonicalized. Fixes 4 tests that fail on any macOS checkout of main.

Testing

• Unit tests for glob semantics (anchoring, depth, dir-only, ! re-includes, invalid-pattern skip, extension normalization).
• End-to-end walker tests: excluded dirs pruned in the workspace scan and the Drupal scan; .module files discovered via extensions.
• Config parse test for the new keys.
• cargo clippy -D warnings, cargo fmt, and the full test suite pass; both commits build and pass independently (bisect-safe).

Refs #48

Checklist

If applicable:

• I have updated CHANGELOG.md
• I have updated the documentation (README.md, docs/, examples/)
• I have updated the config schema (config-schema.json)
• I have added/updated tests to cover my changes
• I fully understand the code that I am submitting (what it does,
  how it works, how it's organized), including any code drafted by an LLM.
• For any prose generated by an LLM, I have proof-read and copy-edited with
  an eye towards deleting anything that is irrelevant, clarifying anything
  that is confusing, and adding details that are relevant. This includes,
  for example, commit descriptions, PR descriptions, and code comments.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 83.70370% with 44 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/indexing/init.rs	22.22%	14 Missing ⚠️
src/util.rs	0.00%	11 Missing ⚠️
src/classmap_scanner/filters.rs	93.96%	7 Missing ⚠️
src/indexing/scan.rs	50.00%	7 Missing ⚠️
src/definition/implementation.rs	0.00%	2 Missing ⚠️
src/analyse/run.rs	83.33%	1 Missing ⚠️
src/fix.rs	0.00%	1 Missing ⚠️
src/indexing/watch.rs	88.88%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!