A Comparative Analysis of Expected and Distributional Reinforcement Learning