Incremental indexing: skip unchanged files + handle deletes

[ryuhmanov-m]

Parent issue: #17

Context

Currently reindex_vault re-embeds every file on every run. The files table already stores SHA-256 hashes, but they're not used to skip unchanged files.

Scope

1. Skip unchanged files

• Before embedding a file, compare its current SHA-256 hash with files.hash in the database
• If unchanged, skip embedding entirely
• Only process new or modified files

2. Handle deleted files

• After scanning the vault, find files in the files table that no longer exist on disk
• Remove their chunks from the chunks table
• Remove the file record from files

3. Handle renamed files

• Detect renames (same hash, different path) — update file_path in chunks instead of re-embedding

Files to look at

• src-tauri/src/adapters/indexer/ — indexing pipeline
• src-tauri/src/adapters/db/ — database queries
• files and chunks table schemas in migrations