Enhancing Deep Reinforcement Learning-based Robot Navigation Generalization through Scenario Augmentation