model: add PyTorch and TensorFlow weights for new Italian ELECTRA model (trained on mC4)

tokenizer: add config and vocab

initial commit