Perform a comparison of the SFM w_2021_28 gen2 vs. gen3 middleware processing runs for the DC2/imsim dataset (i.e. those of DM-30674 & DM-31067) analogous to what was done for the  w_2021_24 on DM-30812 and, in particular, checking on the visit padding hypothesis for the slight WCS differences seen there (this should now be resolved as of PREOPS-580).  Also thanks to PREOPS-580, we should now have the full suite of raws properly ingested and defined in the /repo/rc2 gen3 repo.  A confirmation of this should also be made on this ticket this ensuring the inputs going to the coadds will match for gen2/gen3 (for moving on to those parity comparisons).