Constrained Reinforcement Learning using Distributional Representation for Trustworthy Quadrotor UAV Tracking Control