debug: keep cpu-32-hk pod alive 1h on failure for kubectl exec

Add a final step that sleeps 3600s only when a prior step in the
build-push job fails, so the failed pod stays around long enough to
kubectl exec into and inspect (cert files, buildctl output, etc.)
instead of being torn down immediately.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Author: JavaPythonAIForBAT

.github/workflows/_build_test_image.yaml

@@ -116,3 +116,9 @@ jobs:
             --progress=plain
 
           echo "Image built (not pushed): ${IMAGE_TAG}"
+
+      - name: Keep pod alive for debugging on failure
+        if: failure()
+        run: |
+          echo "A previous step failed — sleeping 1h to allow 'kubectl exec' debugging into this pod."
+          sleep 3600