Factored Adaptation for Non-Stationary Reinforcement Learning