Multi-task Learning and Catastrophic Forgetting in Continual Reinforcement Learning