Estimation and Inference in Distributional Reinforcement Learning