Interpreting Graph Drawing with Multi-Agent Reinforcement Learning