Machine Learning-Driven Attack and Fault Classification in CV-QKD Systems after Anomaly Detection