Integrating Deep Reinforcement and Supervised Learning to Expedite Indoor Mapping