Joint Representations for Reinforcement Learning with Multiple Sensors