feat(memory): add optional subtype filtering

[ramonlimaramos]

Summary

• add nullable memories.subtype plus tenant/subtype index via migration 004
• let remember(..., subtype=...) store optional free-form subcategories
• let recall(..., subtype=...), hybrid_search, and vector_search filter by subtype with parameterized SQL
• share subtype/depth filter construction through a single helper to avoid duplicated search logic
• include subtype in retrieval formatting, export/import JSON, README guidance, and MCP tool descriptions
• add tests for persistence, recall forwarding, subtype search filters, injection safety, and tool metadata

Release plan

• Target: v0.5.0 source-of-truth / governed context layer
• Closes More granular memory_type subcategories with autocompleted enums #51

Review follow-up

• Resolved Clean Code finding from review by extracting duplicate memory filter construction into _build_memory_filters(...).
• Performance note: subtype filtering remains parameterized and O(1) to build in Python; DB-side behavior depends on planner/index selectivity and should be validated with EXPLAIN ANALYZE once representative memory volume exists.

Validation

• git diff --check
• uv run pytest tests/unit/test_migrations.py tests/unit/test_search_and_graph.py tests/unit/test_memory_retrieval_tools.py tests/unit/test_tool_metadata.py -q (70 passed)
• uv run pytest tests/unit/test_search_and_graph.py -q (21 passed after review follow-up)
• uv run pytest tests/unit -q (242 passed after review follow-up)
• uv run ruff check src/synapto/search/hybrid.py tests/unit/test_search_and_graph.py
• uv run ruff check src/ tests/
• GitHub CI run 27515871845: lint, security, test 3.11, test 3.12, test 3.13 all passed