feat: add re-index button and resolve all clippy lints

feat: Remove LLM dependencies and integrate HNSW-based ANN for faster/stable search ...

- Removed llama-server, GGUF models, and heavy LLM DLLs.
- Cleaned up build.rs and tauri.conf.json to lightweight the build.
- Integrated hnsw_rs in mcp.rs for high-performance semantic search.
- Improved LSA robustness for unknown words.
- Reclaimed approx 3GB of workspace space.

feat: implement 800-character document chunking and stabilize Japanese search build ...

- Implement 800-char text chunking in add_item_text.
- Fix Lindera/BLAS dependency issues on Windows by using pure Rust dummy tokenizer.
- Add unit tests for chunking logic in mcp.rs.
- Update .gitignore to include journals.
- Add implementation journal 20260219-0004.

feat: 日本語 LSA セマンティック検索の実装 (Lindera + SVD) ...

- Lindera による日本語形態素解析トークナイザーを追加
- ndarray-linalg / rsvd による LSA (潜在意味解析) コアロジックを実装
- SQLite テーブル (items_lsa) へのベクトル保存機能を追加
- MCP ツール (lsa_search, lsa_retrain) を統合
- 「軽さは正義」に基づき BM25 (FTS5) 依存を排除し、LSA に一本化