Continual Adversarial Reinforcement Learning (CARL) of False Data Injection detection: forgetting and explainability