(When) Are Contrastive Explanations of Reinforcement Learning Helpful?