Features for simple feed-forward and D-Linear

[StefanoDamato]

Very simple architectures, such as feed-forward neural network or D-Linear, can be effective models. However, they are not designed to use additional features (e.g. day of the week, time series ID, etc.). For a personal project I realised an extension which also includes arguments to exploit covariates as num_feat_dynamic_real, num_feat_static_real, num_feat_static_cat, cardinality, and embedding_dimension.

I understand that these features are not part of the original formulation of the models, but they can make them more flexible in scenarios in which additional information is crucial. Would this be a useful contribution to the package? If that is the case, I would be happy to develop and test these modifications.