We need to be able to get AP ZooniverseCutouts using the SourceCatalog Footprints instead of the centroid/shape-based locations. This can help debugging pipeline failures, where e.g. the footprint is on chip but the centroid has been pushed off-chip (see DM-37421).

Relatedly, I need to come up with a formula for making a square bbox from a rectangular one (to turn a non-square footprint into a square one, which simplifies the cutout layout). I don't think that needs any new code, just figuring out the right bbox calls.

Splitting out the afw part of this work (a version of getCutouts that takes pixels and can return something that falls off the chip edge) to DM-38544, as I have a working version of the analysis_ap ZooniverseCutouts code, and will tweak it to use the DM-38544 changes once they're done.