fix: Validate cached MCP indexes on query

[Pringled]

This PR ensures MCP indexes get validated on every query like the CLI route. There's a minimum time (3x original indexing time) to ensure we don't start re-indexing while it's still indexing).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
src/semble/mcp.py	100.00% <100.00%> (ø)
src/semble/version.py	100.00% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[greptile-apps]

Confidence Score: 5/5

Safe to merge — the staleness check is gated behind multiple guards (done task, non-cancelled, no exception, cooldown, identity check before evict) and all mutations to _revalidate_after now happen on the event loop thread.

The core logic is sound: cooldown is recorded after the asyncio.to_thread await (event-loop-thread write), the concurrent-replacement race is handled via identity comparison before eviction, git URLs are bypassed, and failed tasks are left alone. get_validated_cache already accepts str | None for model_path and resolves it internally, so the self._model_path type is compatible. Tests cover all the new code paths including the cooldown gate, failed-task skip, and concurrent-replacement scenario.

No files require special attention.

_{Reviews (2): Last reviewed commit: "Change timing behavior, fix greptile fee..." | Re-trigger Greptile}

[Pringled]

@greptileai review