Improve job deletion

[thomasjo]

It seems there are scenarios where jobs marked for deletion are left in a perpetual state waiting for pods to be terminated. We need to find a way to force the pods (and other transient resources) to be terminated within a grace period or some such.