Upcoming changes to Python library

[alasdairforsythe]

I'm planning the following changes to the Python library which will be implemented within the next couple of days:

1. Tokenize will return numpy array instead of list of ints (decode will accept either).

2. The functions convert_ids_to_tokens, convert_ids_to_tokens_decoded & convert_tokens_to_ids will be removed. I originally intended them for compatibility with Hugging Face Transformers, but it's clear now that was the wrong approach as they cannot be used in the same context as the similarly named functions in Transformer's tokenizer classes. Use token_to_id and id_to_token instead, or preferably decode.

If you have any questions, concerns or suggestions on this please reply here.