Optimizing UAV Trajectory and Power Allocation with Deep Reinforcement Learning for Throughput and Fairness in Full-Duplex Networks