R3HF: Reward Redistribution for Enhancing Reinforcement Learning from Human Feedback