Created by: dilyararimovna
refactor: moved padding to embedder for most configs (except of self attention models, they still need fixed text size), renamed multi_label to single_vector and added it to all classification configs, updated classification example