perf+fix: optimize indexing/search pipeline and harden interface/runtime

[DevvGwardo]

Summary

This PR delivers a broad optimization and reliability pass across Peppy core functions, interface/runtime wiring, and tests.

Core optimizations

• Cache freshness validation (peppy/cache.py)
  • Added manifest-based validation (path, size, mtime) to prevent stale cached indices from being reused after source changes.
• Indexer throughput + search prep (peppy/indexer.py)
  • Added auto worker sizing (max_workers=None => adaptive based on CPU).
  • Added precomputed lookup maps (symbol_lookup, symbol_type_lookup) at index time for faster search queries.
• Search fast paths + pattern semantics (peppy/searcher.py)
  • Added fast-path symbol lookups using precomputed index maps.
  • Normalized file pattern matching to support both relative-to-root and absolute path matching.

Correctness / robustness fixes

• Parser decode safety (peppy/parsers.py)
  • Decode source with errors="replace" in tree-sitter extraction path to avoid failures on imperfect encoding input.
• Regex safety in high-level interface (peppy/interface.py)
  • Escaped user-provided symbol names in find_definition() and find_usages() to prevent regex metacharacter bugs.
• Server async loop quality (peppy/server.py)
  • Switched to asyncio.get_running_loop() for executor dispatch.
• Packaging entrypoint fix (pyproject.toml)
  • Console script now points to sync wrapper (peppy.server:run) instead of async function.

Test improvements

• Unit parser test fix (tests/unit/test_parsers.py)
  • Added missing assert for class name check.
• Performance test setup correction (tests/performance/test_scalability.py)
  • Ensured indexer/searcher share the same cache instance where intended.

Impact

• Better correctness after codebase changes (fewer stale-index surprises).
• Faster repeated symbol queries on medium/large repos.
• More robust behavior with odd file encodings and regex-sensitive names.
• Cleaner runtime behavior and packaging ergonomics.