Is your feature request related to a problem? Please describe.
When a factor-valued treatment is passed to the bcf() function in R, we require that it be binary, in which case it is recoded to 0 / 1. If a user passes a treatment variable as a factor with more than 2 levels, we could alternatively raise a warning that it will be one-hot encoded and fit a multivariate BCF to the recoded treatment.
Describe alternatives you've considered
The alternative is the status quo -- raise an error and require users to recode their factor-valued treatment as an appropriately-coded numeric treatment matrix.