Learning Risk-Aware Quadrupedal Locomotion using Distributional Reinforcement Learning