Feature request: Grouping features

[jemus42]

In conversation with Marvin, rough idea:

• Supply a list of vectors of features to group (currently we can pass features vector only for subselection)
  • example: glex(..., features = list(group1 = c("x1", "x2"), group2 = c("x3", "x4")))
• Groups features into a (named) group
• Use cases:
  • for dummy-encoded variables in XGB, re-aggregating them into one "feature" makes things easier for visualization and interpretation
  • For groups of somewhat related features, e.g. demographic information, lab measurements, socioeconomic features, ...
  • Could also be useful or interesting to collect e.g. x1 and all of its interaction terms (effectively into it's SHAP term) if they are not too interesting while remaining features and their interaction terms remain separated