like podman stop of containers, we should allow the user to specify a timeout override when stopping pods; otherwise they have to wait the full timeout time specified during the pod/container creation. Signed-off-by: baude <bbaude@redhat.com>