Use of `batch_key` argument

[fischer-hub]

Hi, thanks for the nice port!

When looking at the decontx.decontx()docstrings I noticed there is a batch_key argument for processing datasets split by batch. However, I ran into an error when trying to run with this set to one of the .obs variables. Is this part of decontx already implemented? Also, if I was to use the batch_key, wouldn't I have to supply a new clustering for each batch? How would that work with the current implementation?

Thanks in advance for the help!