Statistical Inference in Reinforcement Learning: A Selective Survey