DM-11023: CoaddPsf: add more exception info

[PaulPrice]

A common problem when calling CoaddPsf.computePsfImage() is that
WarpedPsf.computeBBoxFromTransform throws a RangeError stating
'Unexpectedly large transform passed to WarpedPsf'. This is
usually because the astrometry for an input is poor, but it would
be nice to identify which input: so add the exposure ID to the
exception message.

[SimonKrughoff]

Two things.
First thing, is it the right thing to do to Except here? I feel like it would be o.k. to report the failure (with dataId) drop it and continue on. The current system makes astrometry failures separated greatly from this piece of code completely halt the processing. It would be nice to be able to continue on even if a few failures are encountered and fix up the failures if necessary. What do you think?
Second thing, is this code covered by a unit test? It would be nice if it was.

[PaulPrice]

I don't see how it's possible to continue on. We could ignore the error and not include that image's Psf in the CoaddPsf, but that means the CoaddPsf would be wrong. Or we could remove the test in WarpedPsf.computeBBoxFromTransform that's throwing the exception in the first place, but that means we're going to have to get even further along before discovering something's bad. I think what we need is better quality control on the astrometry (as you suggest, identify problems close to where the problem occurs), and I understand that's something that's being actively worked on so it's not something I intend to tackle here.
Regarding a unit test, I would prefer to avoid testing the contents of an exception message, as it's something that's not so important and can change easily (line numbers will change, additional information may be added) --- the same reasons that we don't test the exact contents of log outputs.

[SimonKrughoff]

O.K. fair point with regard to discovering astrometric problems early.
Regarding, the unit test, I was not asking to test the content of the exception, just that an exception gets thrown. I'll leave it up to you, but it would be nice if this code were at least called in a unit test somewhere.

[SimonKrughoff]

See comments in the general PR discussion.

[PaulPrice]

OK, I just checked and I don't see an existing test of the exception. Adding one is more work than I originally signed up for, but I'll see if I can make one quickly.

[PaulPrice]

Added a test, and rebased against master.