Widening the Pipeline in Human-Guided Reinforcement Learning with Explanation and Context-Aware Data Augmentation