[FEA] Record Leiden/Louvain modularity in `uns`

[jpintar]

Right now, the modularity achieved in Leiden clustering is being thrown away here:

rapids_singlecell/src/rapids_singlecell/tools/_clustering.py

Line 156 in 4e8ac31

leiden_parts, _ = culeiden(

(and similarly for Louvain further down in line 292).

Would you consider recording it in adata.uns["leiden"] instead?

I am aware that there will eventually be a scanpy.metrics.modularity function, which might get a corresponding implementation here, but as I comment in the associated discussion, saving the already computed modularity is basically free, and could potentially be quite useful (e.g. when one is computing multiple clusterings of the same data, perhaps varying the resolution parameter, or the number of nearest neighbours, and comparing the results).

I'd be happy to write a small PR for this, if you agree.