Unsupervised Representation Learning in Deep Reinforcement Learning: A Review