Taking phenotype column in diff_nhood_enrichment

[LukasHats]

Hey @marcovarrone,

just wanted to get your opinion on the 'robustness/validity' on this approach:
If we take the Phenotype .obs as input for cc.pl.diff_nhood_enrichment instead of the neighborhood column, are we essentially doing a classic co-localization/interaction analysis? What do you think, is the function suited for that or are there any caveats that might be problematic? Because your function is so handy for comparing cohorts, I would love to also try it on phenotype labels.

Best,
Lukas