The cluster's API server rejects the limits field in Deployment containers. Keep only requests (CPU/memory) for now.
Namespace and ConfigMap are managed manually (not by Helm).