Implement integration benchmarking using scIB

[nictru]

Description of feature

Users should be able to systematically evaluate the quality of the created integrations. scIB seems like the most suitable tool for this. Has already been implemented in SIMBA. Results should be added to the MultiQC report.

[ryan2han]

When using the integration benchmark with scIB, it requires a label_key. Will users need to prepare their own annotations in advance, or can we utilize the cell-type labels from celltypist as the label_key?

[nictru]

I think both options can easily be implemented, so I will probably add a parameter for selection.

[nictru]

Can be based on the implemenation from here