Reinforcement Learning with Fast and Forgetful Memory