Implement batch key argument in sc.pp.highly_variable_genes

[crichgriffin]

Wanted to implement the batch key argument, left default as None so backwards compatability is maintained
See sc.pp.highly_variable genes for details about the argument

[crichgriffin]

@deevdevil88 suggested:
Allow for the HVG batch key to account for multiple covariates by combining them into one column as we do for running the integration methods.
I agree and I will implement this

[deevdevil88]

Hiya!
I had a minor update suggestion, it was just, that in the integration workflow, for batch correction, there is an option to combine two columns or covariates to make a joint batch_key, i was wondering since, you are implementing batch_key for HVG, if this can also be done for this step as well. As otherwise for this step, if we want to combine two columns for our batch, we would need to do it manually (which isnt the end of the world) but it would just be consistent with the integration step to make the option available

[crichgriffin]

I have address @deevdevil88 comments.

I also replaced all the columns = [x.replace(" ", "") for x in args.integration_col.split(",")] with columns = [x.strip() for x in args.integration_col.split(",")]in all the batch correction scripts, as the strip function only removes trailing spaces, whereas the replace function would have removed spaces within column names. Spaces within column names are not advisable, but are allowed by pandas and anndata.