Loss of Plasticity in Continual Deep Reinforcement Learning