Exploring Data Scaling Trends and Effects in Reinforcement Learning from Human Feedback