Added the doc om MLP training using AE embeddings

This MR is the final block of documentation related to training of autoencoders-based multi-channel face PAD. Doesn't contain any PyTorch related configs, just in the documentation.