We don't need to make the Monte Carlo Search Tree object live inside of the trainer. It can exist as a local inside `execute_episode()`.