Hybrid-AIRL: Enhancing Inverse Reinforcement Learning with Supervised Expert Guidance