perf: optimize core search/index functions for Claude workflows

[DevvGwardo]

Summary

This PR applies targeted performance and correctness improvements across the core indexing/search stack used by Claude workflows.

Changes

peppy/searcher.py

• Added max_results support to search_symbols() with early termination to avoid full index scans.
• Improved file-pattern matching against relative path, basename, and full path.
• Fixed get_file_symbols() relative path handling by resolving relative inputs from codebase root.
• Reduced grep overhead and only builds context payload when requested.

peppy/interface.py

• find_symbols() now passes limit down to search layer (max_results) so limiting happens during search, not after.
• Tightened get_overview() entrypoint searches with narrower patterns and smaller per-pattern caps.
• Reuses precomputed symbol/file-extension stats when available.

peppy/server.py

• Extended search_symbols MCP schema with max_results.
• Wired max_results through tool execution into searcher for backend compute control.

peppy/parsers.py

• Fixed parser traversal context leakage: class context is now scoped per branch so sibling nodes are not misclassified.

peppy/cache.py

• Upgraded cache key hashing to sha256.
• Added optional TTL support (ttl_seconds).
• Added freshness signature validation to avoid stale index reuse when indexed files change.

Why

• Reduce unnecessary work on large repositories.
• Improve token/time efficiency for planning and symbol discovery in Claude-assisted workflows.
• Improve correctness (symbol typing + file lookup + stale cache behavior).

Compatibility

• Backward compatible API changes (new optional args).
• Existing workflows continue to work with improved performance/correctness.