Use wait --for=delete on pods instead of jsonpath on deployment

Kubernetes omits .status.replicas entirely when it reaches 0
rather than setting it to 0, so the jsonpath condition never
matches and the wait times out.

Wait for the actual pods to be deleted instead, which directly
confirms the Go processes have exited and flushed coverage data.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: pedjak

hack/test/e2e-coverage.sh

@@ -22,9 +22,9 @@ rm -rf ${COVERAGE_DIR} && mkdir -p ${COVERAGE_DIR}
 kubectl -n "$OPERATOR_CONTROLLER_NAMESPACE" scale deployment/"$OPERATOR_CONTROLLER_MANAGER_DEPLOYMENT_NAME" --replicas=0
 kubectl -n "$CATALOGD_NAMESPACE" scale deployment/"$CATALOGD_MANAGER_DEPLOYMENT_NAME" --replicas=0
 
-# Wait for deployments to scale down so coverage data is flushed to the PVC
-kubectl -n "$OPERATOR_CONTROLLER_NAMESPACE" wait --for=jsonpath='{.status.replicas}'=0 deployment/"$OPERATOR_CONTROLLER_MANAGER_DEPLOYMENT_NAME" --timeout=60s
-kubectl -n "$CATALOGD_NAMESPACE" wait --for=jsonpath='{.status.replicas}'=0 deployment/"$CATALOGD_MANAGER_DEPLOYMENT_NAME" --timeout=60s
+# Wait for manager pods to terminate so coverage data is flushed to the PVC
+kubectl -n "$OPERATOR_CONTROLLER_NAMESPACE" wait --for=delete pods -l control-plane="$OPERATOR_CONTROLLER_MANAGER_DEPLOYMENT_NAME" --timeout=60s
+kubectl -n "$CATALOGD_NAMESPACE" wait --for=delete pods -l control-plane="$CATALOGD_MANAGER_DEPLOYMENT_NAME" --timeout=60s
 
 # Copy the coverage data from the temporary pod
 kubectl -n "$OPERATOR_CONTROLLER_NAMESPACE" cp "$COPY_POD_NAME":/e2e-coverage/ "$COVERAGE_DIR"