Combining sequential Gaussian co-simulation and Monte Carlo dropout-based deep learning models for geochemical anomaly detection and uncertainty assessment