Towards Robust Policy: Enhancing Offline Reinforcement Learning with Adversarial Attacks and Defenses