Allow block cross-validation classes to take extra coordinates

[leouieda]

Description of the desired feature:

Sometimes, we may want to pass extra variables to be split (like heights or other numerical/categorical values) to be split by BlockKFold and BlockShuffleSplit (see fatiando/community#86). Right now, there are two things blocking this:

1. The BaseBlockCrossValidator checks for 2 dimensions on X: https://github.com/fatiando/verde/blob/main/verde/base/base_classes.py#L81
2. The cross-validators assume easting=X[:,0] and northing=X[:,1]: https://github.com/fatiando/verde/blob/main/verde/model_selection.py#L184

A solution would be to:

1. Remove the check in the base class.
2. Pass in an argument coordinate_columns=[0, 1] that specifies which columns of the feature matrix X correspond to the coordinates.

I think the rest should work and make these classes more usable to general spatial machine learning.

Are you willing to help implement and maintain this feature?

Maybe but I have no bandwidth for this right now and don't really have a use for it just yet. Happy for anyone else to give it a shot.