Support categorical/oblique split and categorical/pivot/proximity splits

[acascione]

For now, PivotTree splits and their variants, along with Proximity Tree splits, are supported only for datasets with numeric variable. We plan to extend it to handle categorical variables and mixed cases.

For Oblique Splits, they handle numeric values but it would be optimal to allow the selection of categorical splits during training in case a mixed variables dataset is provided and a categorical univariate split is optimal with respect to other multivariate continuous variable-based splits