bug: fix OpenShift deploy script CatalogSource and CSV wait

The CatalogSource referenced a non-existent "latest" image tag,
causing ImagePullBackOff. The CI publishes the catalog image with
the git ref name (e.g., "main"), so use that tag instead.

The CSV wait loop used items[0] which matched the first CSV
alphabetically (e.g., authorino-operator) rather than mcp-gateway,
causing the script to proceed before CRDs were established. Use
OLM's operator label selector to target the correct CSV.

Signed-off-by: Matthias Wessendorf <mwessend@redhat.com>

Author: matzew

config/deploy/olm/catalogsource.yaml

@@ -4,7 +4,7 @@ metadata:
   name: mcp-gateway-catalog
 spec:
   sourceType: grpc
-  image: ghcr.io/kuadrant/mcp-controller-catalog:v0.5.1-rc1
+  image: ghcr.io/kuadrant/mcp-controller-catalog:main
   displayName: MCP Gateway
   grpcPodConfig:
     securityContextConfig: restricted

config/openshift/deploy_openshift.sh

@@ -70,7 +70,7 @@ kubectl apply -f /tmp/mcp-subscription.yaml -n $MCP_GATEWAY_NAMESPACE
 
 echo "Waiting for controller CSV to succeed..."
 retries=0
-until kubectl get csv -n $MCP_GATEWAY_NAMESPACE -o jsonpath='{.items[0].status.phase}' 2>/dev/null | grep -q "Succeeded"; do
+until kubectl get csv -n $MCP_GATEWAY_NAMESPACE -l operators.coreos.com/mcp-gateway.$MCP_GATEWAY_NAMESPACE="" -o jsonpath='{.items[0].status.phase}' 2>/dev/null | grep -q "Succeeded"; do
   retries=$((retries + 1))
   if [ $retries -ge 60 ]; then
     echo "Timed out waiting for controller CSV to succeed"