[BUG] Restarting a run leaves GcpInstance, so CI never re-queues it #1224

Description

restart_run (API) and restart_test (UI) clear TestProgress / results so the run looks pending again. They do not delete the GcpInstance row.

gcp_instance / start_test treat any test with a GcpInstance as already running and skip it. After restarting a running or stuck run, cron therefore never starts a new VM.

Expected

Restart clears the instance reservation so the same test id can be picked up again. The old VM, if still present, becomes orphaned and can be cleaned up by the existing orphan scanner.

Actual

The GcpInstance row stays. start_test logs that the test already has a GCP instance and skips starting it.

Fix

Delete GcpInstance rows for that test_id in both restart paths.

Tests

Regression coverage on API and UI restart when an instance row exists.