Use of padding index in tokenizer.tokenize + utils.tokenized_text_in_tokens function

[jpramil]

Might be an issue with setting a default value for the padding index in the utils.tokenized_text_in_tokens function.
Actually, more specifically, I think this function is only used in the tokenizer.tokenize function, which in turn is only used in the predict method of the FastTextModel class.
But it is called before the actual padding index is calculated (which depends on the input data).