Is your feature request related to a problem? Please describe.
Currently, CausalTreeRegressor doesn't support multiple treatments, unlike UpliftTreeClassifier or meta-learners.

Describe the solution you'd like
We can follow the implementation in UpliftTreeClassifier, which evaluates splitting criteria across all treatment groups at each node.

Describe alternatives you've considered
Training one CausalTreeRegressor per each treatment separately, then combining the results. This will be a quick and dirty solution.

Additional context