Improving Human Sequential Decision-Making with Reinforcement Learning