Consider warning users when using relative abundances with scCODA

[maltekuehl]

Description of feature

Somebody (definitely not me) might have naively used relative sample-level abundances (columns summing to 1) as inpus to scCODA and gotten very confused at log2FCs of 0 for all variables.

Having a small warning that absolute abundances are required might prevent others (again, definitely not me) from losing half an hour debugging. Should be straightforward to detect. In prepare (https://github.com/scverse/pertpy/blob/main/pertpy/tools/_coda/_base_coda.py#L168), we have:

sample_adata.obsm["sample_counts"] = np.sum(sample_adata.X, axis=1)

This could be checked to be > 1, if I understand the code correctly.

Thanks!

[Zethson]

Thanks! I'll take care of this soon.