Add the existing instructions for debugging API CPU and memory usage to the docs. This should cover selecting the cluster context, port-forwarding the API service, collecting heap and CPU profiles from the pprof endpoints, and opening the resulting profile with go tool pprof.