Robustness and Generalization in Quantum Reinforcement Learning via Lipschitz Regularization