Adding `safetensors` variant of this model (#1)

readme: add initial version

models: add PyTorch and TensorFlow versions

tokenizer: add config (no accent stripping) and vocab

config: add initial version

initial commit