An experimental design perspective on model-based reinforcement learning