[R] Throw explicit error when using categorical features in functions which don't support them

[david-cortes]

ref #9810
ref #9925

This PR adds an explicit error message when trying to call xgb.model.dt.tree and functions that depend on it (currently xgb.plot.multi.trees) on a model that has categorical features, as it fails to extract the splits from those.

It doesn't actually fix the issue - i.e. this PR doesn't fix the functions to make them work with categorical data - but instead just makes the error message understandable.

Should ideally be reverted once there is a C-level function to produce tables from trees with proper support for categorical splits.

[trivialfis]

Will merge once the CI is green.

[david-cortes]

On second thought, I'm thinking it better to remove the text argument for trees-to-table:

• Once a C-level function to get the table is added, it'd work from a booster object only instead of text dumps, so better to have the breaking change now before a CRAN release.
• Having this text argument doesn't allow throwing an informative error when the model has categorical features.

[trivialfis]

Once a C-level function to get the table is added

It might take a while, though. I'm working on storing the factor in the booster, as there are complaints about this being a foot gun.

[trivialfis]

Could you please help resolve the conflicts?

[david-cortes]

Could you please help resolve the conflicts?

Solved.