Continual Model-Based Reinforcement Learning with Hypernetworks