RAG-Reward: Optimizing RAG with Reward Modeling and RLHF