Training Reinforcement Learning Agents and Humans With Difficulty-Conditioned Generators