Differential Privacy for Anomaly Detection: Analyzing the Trade-off Between Privacy and Explainability