Shutdown during startup no longer reports a database failure

At startup the portal runs a gate that checks the database schema is fully migrated before it accepts traffic. When the portal was asked to shut down while that check was still talking to the database — for example a deployment rollout replacing a pod moments after it started — the interrupted check was reported as a critical "DB version check failed unexpectedly. Refusing to start the portal." Nothing had failed: the portal was simply told to stop, and the check was called off with it.

An interrupted startup check now stands down quietly, the way the earlier fix already taught the runtime health probe to treat its cancellations. Genuine startup problems — an unreachable database, a missing schema, an incomplete migration — are still reported at critical level and still stop the portal from starting, exactly as before.