Curiosity-Driven Reinforcement Learning from Human Feedback CD-RLHF: An AI Framework that Mitigates the Diversity Alignment Trade-off In Language Models