Tiny compat fix: sklearn expects fit_transform to take a y as well

[Slater-Victoroff]

Hi Daniel! Apologies for not getting back to you on this sooner, but my understanding is that while the goal of passage is to extend an interface very similar to sklearn for the sake of usability, there are certain areas where we're okay with breaking the standard in order to make the interfaces more intuitive.

In this case I think we're going to leave out the extra dummy parameter despite the sklearn compliance to keep things more clear.

[dnouri]

No worries about the late reply.

Currently, without the dummy y parameter, it's impossible to use the Tokenizer as part of a sklearn Pipeline. I use Pipelines all the time and would like to see Tokenizer work with them.