Weight Clipping for Deep Continual and Reinforcement Learning