Reinforcement Learning with Non-Exponential Discounting