The Cell Must Go On: Agar.io for Continual Reinforcement Learning