Learning Lower Bounds for Graph Exploration With Reinforcement Learning