Use _convert_id_to_token for create tokenizer vocab dict

[jumelet]

Instead of having to create a separate mapping for each model class, each tokenizer seems to have implemented its own version of _convert_id_to_token that can be used directly without overhead.