Add vocabulary dumped from various tokenizers together with processing scripts

initial commit