Refactor BERTTokenizer

[mhabedank]

The BERTTokenizer is using torchtext. We want to remove torchtext as a dependency so this Tokenizer has to be refactored not using it.

[ManishSharma1609]

Hi, I’m interested in contributing a fix for this issue if it would be welcome. I’d appreciate some guidance on how to approach the refactoring and any specific requirements to keep in mind.