GAN-Based Interactive Reinforcement Learning from Demonstration and Human Evaluative Feedback