What Hides behind Unfairness? Exploring Dynamics Fairness in Reinforcement Learning