Improve error reporting on kubectl failure

When a kubectl command fails we do not consistently report any captured stderr in the exception.
e.g. there a few places that only report "Failed to replace k8s resource" as the exception, which means the underlying cause/issue is not reported (only logged).