Action Gaps and Advantages in Continuous-Time Distributional Reinforcement Learning