feat(cli): show last anomaly in status (#91)

[bferanmi806-sketch]

Closes #91.

What changed: ProfileStore gains get_latest_anomaly_per_table(), one window-function query returning the latest anomaly per table (highest detected_at, id tie-break, no 30-day window). status() calls it once and attaches last_anomaly ({type, severity, detected_at} or null) to each record. Text output gains a Last Anomaly column ( · , e.g. volume_drop · 2h ago, empty when none); JSON keeps all existing keys. _table_health, persistence, schemas, and connectors untouched.

Tests: pytest tests/test_store.py tests/test_cli.py — 41 passed. Full pytest — 457 passed; 8 failures in test_demo/test_duckdb/test_mcp_server are pre-existing on clean main (verified via stash: identical 8 fail without this change; duckdb/MCP environment issues). ruff check src/ tests/ — all checks passed.

[rbmuller]

Tested it end to end against scherlok demo data: text and JSON both look right, one query for all tables, and the severity comes through clean. Merging, thanks.

On the 8 failures you mentioned: they don't reproduce here — main is green in CI and locally (465 passed on your branch), so it's likely the duckdb/mcp install in your environment rather than the code.

One small follow-up I'll track separately: in the text table, OK next to volume_drop · just now can read as a contradiction, since Status is a live check and Last Anomaly is history. Showing the severity in that cell would make it clearer.