v0.5.0

Added

• Quoted Phrase Search: Use quotes in mneme find to search for exact phrases (e.g., mneme find "error handling" go). Multi-word quoted arguments are treated as phrase matches in snippet highlighting, while individual words are matched separately.
• Query Parsing Module: New internal/query/parse.go with ParseQueryInput for structured tokenization of mixed phrase and keyword queries, with full test coverage.
• Benchmark Suite: New internal/benchmark package with a build-tagged (//go:build benchmark) test suite that measures crawl, index, search, read, and write throughput across corpus sizes of 100–5,000 files. Results are printed as a formatted table. Run via make benchmarks.
• TESTING.md: Added documentation for the benchmark targets.

Changed

• Find Command UX: Improved Long description and Example fields for the find command with clear usage patterns for phrase vs. keyword search.
• Search Result Filtering: Results are now filtered to only include documents with actual text-matching snippets, reducing false positives from BM25 stemming.
• Code Cleanup: Removed unused exported functions, aligned comments, and applied CodeRabbit review suggestions across multiple packages.

Fixed

• Benchmark Isolation: Fixed benchmark tests leaking into the user's real config and data directories by saving and restoring constants.ConfigPath alongside constants.DirPath.
• Test Stability: Fixed failing tests caused by stray quote characters in query argument handling.