store y_var in fit output from xy interface when available

[simonpcouch]

Related to tidymodels/workflows#131. That issue notes that augment.workflow() doesn't return .resid as does the equivalent model fit from augment.model_fit(), even though it calls the model_fit methods in its internals. This is because workflows always(?) transitions to the xy interface before handing data off to parsnip. I came across this while putting together the companion PR to #960.

While the xy machinery makes use of the ..y placeholder in its internals, we can append the actual names attached to y—that are, at least in the workflows case, still intact—as y_var to the resulting model_fit so that we can reference that column later.

In combination with a workflows PR that will follow up on this one, augment.workflow() will return residuals in the situations that augment.model_fit() does!

[hfrick]

I looked at this more from a "pure parsnip" perspective, to complement your workflow lense and found two use cases that you are unlikely/not going to run into when fitting through workflows but should get some attention anyways. Other than that: 👍 😄

And just as an aside:
re workflows "always(?)" transitioning to the xy interface: IIRC, it goes through fit() if the workflow has a "model formula" aka something in the formula arg to add_model(). That's used when you need to specify specific terms or use specific syntax that say a recipe can't understand. Examples are strata() terms for survival analysis or random effects in hierarchical models.

[simonpcouch]

## revdepcheck results

We checked 56 reverse dependencies, comparing R CMD check results across CRAN and dev versions of this package.

 * We saw 0 new problems
 * We failed to check 0 packages

🚀

[simonpcouch]

Deferring to you on merge of this PR, @topepo!

[github-actions]

This pull request has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.