Towards an Interpretable Autoencoder: A Decision-Tree-Based Autoencoder and its Application in Anomaly Detection