tuning multiplexer post

[jakob-r]

This post is losely based on the following issue by @missuse https://github.com/mlr-org/mlr3book/issues/116

It covers the following aspects:

• Build a pipeline that can switch between multiple learners.
• Define the hyperparameter search space for the pipeline.
  • Define transformations for single hyperparameters.
  • Define a hierarchical order of the hyperparameters.
• Run a random search.

[berndbischl]

this is a super important usecase and post. thanks. i couldnt review every line yet, but this should be merged soon. even if we improve it later

[missuse]

Good thing I did not attempt to transform (trafo) the hyper parameters in my example. I would probably still be working out what went wrong.