Supporting quantile regression with XGBoost

[pbhogale]

Feature

In situations when one wants to use R to do a quantile regression, the options available are fairly limited - the quantreg package and quantregForests are two.

On the other hand, since version 2.0.0, XGBoost also provides a quantile regression option, available in the R package as well.

1. It would be useful in and off itself to provide an interface to this capability from boost_tree so that quantile regression with xgboost is available out of the box
2. The function probably::int_conformal_quantile it might also be useful to give an option to use the xgboost quantile regression or the quantreg rq function instead of regression forests.

[joranE]

I also am extremely interested in quantile regression being added to parsnip, but that feature is (AFAICT) not available yet in the xgboost R package, and probably will be some time coming still. They are rewriting the entire R interface to the underlying xgboost library and won't enable quantile regression until that's done. Skimming the linked issue, I think there is still a lot of work to do.