Resolve "Gapfilling + MLP model"
Closes #6 (closed)
I am not happy with the contorsions needed to feed the data from the dataloader to the SITS formalism (and the shape of the mask and the dates tensors expected by the linear gapfillin). See https://src.koda.cnrs.fr/mmdc/mtan_s1s2_classif/-/blob/6-gapfilling-mlp/tests/test_gapfilling_mlp.py?ref_type=heads#L80.
Maybe we should modify the data loader, but maybe it's better to see how the data is used with other models?