Importing `recipes` introduces many implicit dependencies

[echasnovski]

At current moment, installation of rsample (supposing tidyverse is installed) causes many packages to be installed as dependencies of recipes. I had 20 extra packages installed (excluding recipes itself):

List of extra packages

‘numDeriv’, ‘SQUAREM’, ‘abind’, ‘lava’, ‘kernlab’, ‘CVST’, ‘DEoptimR’, ‘magic’, ‘prodlim’, ‘DRR’, ‘robustbase’, ‘sfsmisc’, ‘geometry’, ‘ipred’, ‘dimRed’, ‘timeDate’, ‘ddalpha’, ‘gower’, ‘RcppRoll’, ‘pls’

I understand the reason to facilitate using the recipes with rsample by providing prepper() wrapper, but I think 20 packages not explicitly needed for core functionality is a little bit too much for one simple wrapper. Maybe it is a good idea to move recipes back to Suggests, mention it with wrapper in vignette, and possibly add rsample into recipes Import (as it already has many imported packages)? Keep rsample more light-weight?

Also current situation might be inconvenient for package authors willing to import rsample.

[topepo]

It would make sense to move prepper to recipes (which would import rsample). I'll work on that.

[ClaytonJY]

@topepo would it also make sense to move form_pred to recipes? Seems a bit out-of-place here and doesn't seem to be used internally at all.

[topepo]

I'm going to leave that one in rsample. You don't really need it for recipes and it is just a little function to help with training materials.