Added a decorator to force transform_output="pandas" when calling scBoolSeq().fit(X) / scBoolSeq().transform(X) / scBoolSeq().binarize(X).
This means that scBoolSeq can now be effortlessly integrated in scRNA-seq pipelines currently using scanpy, without using with sklearn.config_context(transform_output="pandas").
Parallel processing is now performed using sklearn.model_selection.KFold(n_workers, shuffle=False).split(df), rather than numpy.array_split(df) which would raise a FutureWarning, and won't be fixed as explained in numpy issue #24889.