How does Kubernetes calculate the maximum number of pods? Does it consider only running pods or completed pods as well?