Specify features and target when creating a `TaggedTable`

[lars-reimann]

Is your feature request related to a problem? Please describe

At the moment we only specify the target when we create a TaggedTable. Based on this, we split the columns into the two categories features, and target.

However, some columns should neither be used to make a prediction (feature) nor be predicted (target) but are necessary understand the predictions. Example: In the Titanic dataset the name of a person is helpful when viewing the predicted survival.

Desired solution

When creating a TaggedTable, specify feature columns and the target column explicitly. We could make the parameter for feature columns optional and keep the default behavior that everything that is not a target column becomes a feature column.

[lars-reimann]

We should also change the constructor of TaggedTable to be similar to the constructor of Table. Additionally, we can offer a method to_tagged_table on Table or a method from_table on TaggedTable. to_tagged_table is probably nicer to work with since it allows method chaining.

Other name: tag_columns - We still get a Table back from the method, so it's not a conversion to something else (so to_XY doesn't make much sense).