Try out cameraGeom+boresight+rotation approach to initial WCSs

We're hoping to start attaching improved initial WCSs using cameraGeom + boresight and rotation (DM-20154), but this depends on the quality of:

• the cameraGeom models
• the boresights from the raw headers
• the rotation angles (when appropriate) from the headers

Looking at the "easy" dataset Lauren MacArthur has been collecting, pick one full visit from each of (HSC, DECam, CFHT, LSST/DC2), and:

• Process the data through at least CalibrateTask (so we can consider those fitted WCSs "truth", at least for now, because "easy" means we should be getting good fits).
• For each detector, load the raw and apply the new function to create the from-raw WCS that John Parejko will provide as a Python snippet.
• Compare the fitted WCS and from-raw WCS.  I recommend starting by looking at how the calexp bounding boxes transform onto the sky (the notebook attached to DM-19731 does this for HSC and may be useful as a starting point).  But looking at other points is of course fine, too, and would avoid some rotational degeneracies.

We're interested in the following questions:

• Does the from-raw WCS function not run at all for one of the cameras of interest, because:
  • the Visitinfo doesn't include boresight or rotation?
  • the cameraGeom lacks a coordinate system we need?
• If the function appears to work but produces garbage, is it because:
  • the VisitInfo boresight values are bad?
  • the Visitinfo rotation values are bad (or perhaps inconsistently defined across cameras)?
  • the cameraGeom does not attempt to include optical distortion at all?
  • the cameraGeom has distortion, but it's not very good?
  • all of the above appear in order, but there's a scaling problem aligned with the zenith direction (i.e atmospheric refraction, which this model doesn't account for but perhaps could)?

I think I already know that the answer for HSC will be "it appears to work but produces garbage because the cameraGeom distortion has a units problem".  It would be nice to hack something together to address that problem, in order to see if there are other problems, but I haven't thought much about how best to do that.