feat(memory): add SQLite FTS5 search index and memory_search tool (#616)

[cagdasyurekli]

Closes #616

Summary

This PR implements full-text search indexing for SQLiteMemoryStore using SQLite's built-in FTS5 extension:

• Creates companion virtual table memories_fts with external content table semantics (content=memories, content_rowid=id).
• Keeps memories_fts automatically synchronized with memories via SQLite AFTER INSERT, AFTER DELETE, and AFTER UPDATE triggers.
• Automatically rebuilds the FTS index when initializing over an existing database with pre-existing memories.
• Adds an abstract search() method to MemoryStore and implements it in SQLiteMemoryStore with tokenized prefix matching, rank ordering, and scope/workspace filtering (matching global memories and current workspace memories).
• Implements graceful fallback to LIKE matching if SQLite lacks FTS5 or encounters operational errors.
• Exposes memory_search(query: str, limit: int = 5) tool through memory_tools() to allow agents to search relevant facts on demand.
• Adds comprehensive unit test coverage in tests/test_memory.py.

Verification

• pytest tests/test_memory.py tests/test_projects.py (64/64 passed).