Stopped-cluster external job-list gate

Motivation

A scoped GET /api/blast/jobs request still resolved the Kubernetes Service and called the external OpenAPI plane while the selected AKS cluster was stopped. The route returned HTTP 200 with local Table rows, but the handled DNS failure created an Application Insights exception and added avoidable latency. The subscription-wide discovery path already excluded stopped clusters; only the explicit-cluster branch lacked the same guard.

Change

The external-jobs service now consults the existing cached ARM cluster-health gate before resolving an explicitly selected OpenAPI endpoint. A proven stopped or missing cluster returns no external target, while local persisted jobs remain available. Missing scope or an unavailable ARM plane preserves the existing degrade-open behavior.

No API schema, persisted row, job state, or running-cluster behavior changed.

Validation

• uv run pytest -q api/tests/test_external_blast_api.py -k 'external_jobs_target or discover_subscription_clusters or collect_and_sync_external_jobs' - 5 passed.
• Production evidence: the affected cluster reported Stopped/Succeeded; the correlated /api/blast/jobs request returned HTTP 200 while its unnecessary service lookup emitted the handled DNS exception.