Reinforcing Thinking through Reasoning-Enhanced Reward Models