Pseudobulk improvements

[Lilly-May]

PR Checklist

• Referenced issue is linked (closes Pseudobulk space improvements #501)
• Documentation in docs is updated

Description of changes

• Added a parameter explanation for the groups_col parameter. There are use cases where one might want to specify groups to consider in addition to the perturbation. That way, one can, for example, summarize perturbations for each cell type individually, so that the resulting pseudobulk object would have separate datapoints for pert1_celltype1, pert1_celltype2, and so on. However, the default setting for this parameter should be None, and the parameter should be explained in the docs.
• Deleted groups_col parameter from the docs example (previously, groups_col was set to the same value as target_col, which is useless since decoupler then internally sets groups_col to None again, see code here).
• Added cast for .obs column storing the perturbation labels to category type

[Zethson]

Thank you!