Improved Sample Complexity Bounds for Distributionally Robust Reinforcement Learning