Add TSFM feature catalog store - #463

Summary

• Cherry-pick and scope PR Tsfm revision #440 feature catalog store code onto current main.
• Remove the legacy TSFM forecasting, fine-tuning, anomaly, model-artifact, and heavy ML runtime tools.
• Add a TSFM feature catalog server with list/search/get/register/update/deprecate/version/lineage MCP tools.
• Add database-backed and in-memory feature catalog lifecycle helpers for transform and extractor cards.
• Read the runtime feature catalog database name from FEATURE_CATALOG_DBNAME, defaulting to feature_catalog.
• Add IoT-style Args/Returns docstrings, successful response message fields, and generic database/catalog wording for TSFM tools.
• Update README, setup docs, and MCP server docs for the current TSFM feature catalog surface.
• Add focused in-memory tests for store behavior, runtime collection selection, and MCP tool wiring.

Tests

• uv run pytest src/servers/tsfm/tests -q
• python -m py_compile src/servers/tsfm/main.py src/servers/tsfm/models.py src/servers/tsfm/core/schemas.py src/servers/tsfm/core/store.py src/servers/tsfm/engine/feature_runner.py src/servers/tsfm/stores/feature_store.py
• rg -n "get_ai_tasks|get_tsfm_models|run_tsfm_forecasting|run_tsfm_finetuning|run_tsad|run_integrated_tsad|PATH_TO_MODELS_DIR|PATH_TO_DATASETS_DIR" README.md INSTRUCTIONS.md docs -S