Make data augmentation available for deep learning
When using Pytorch, data augmentation is not possible because it uses a sqlite file not created in the case of deep learning (a netCDF file replaces it). We suggest creating this file when data augmentation is requested by the user, then performing the data augmentation and generating the expected netCDF file for the rest of the workflow. More over, it doesn't seem appropriate to perform data augmentation on non-gapfilled data, so we suggest adding a check to prevent the use of data augmentation when gapfilling is disabled.