RICE: Breaking Through the Training Bottlenecks of Reinforcement Learning with Explanation