DiffPoGAN: Diffusion Policies with Generative Adversarial Networks for Offline Reinforcement Learning