Rethinking the Foundations for Continual Reinforcement Learning