Robotic Auxiliary Losses for continuous reinforcement learning