Only read the calexps when building the ccdImages (and not when
building the output to write) and save the Calib to its respective
ccdImage to use it when preparing output. Use that calib in ccdImage
to convert flux to magnitude, instead of custom code.

No need for the calexp_md now: just extract the values from the calexp
(eventually this should come from some metadata datasetType, e.g.
`get('info')`).

Add SOURCE_IO_NO_FOOTPRINTS flag to get('src') for a big speedup,
since I don't need the heavy footprints (thanks @TallJimbo for the
suggestion).

Add individual profiling for each slow step, activated via
profile_jointcal option.

Remove a bunch of variables and getters from ccdImage that are totally
unused.

With plot_jointcal_results, I can now extract statistics from any
jointcal run, so I don't need those tests sitting around here.

Split photo/astro code with Config parameters, and a test for each option in the
lsstSim test class. The tests check for empty Wcs/Calib, if they shouldn't have
been written.

Pulled apart _testJointcalTask so I could use the run/plot bits in the "only
one" tests, since those have quite different assert conditions.

Also support the split in JointcalStatistics: the nested ifs are a bit
excessive, but they get the job done.

Fixed inconsistent capitalization of jointcal in tests.

Add astrometry/photometryRefObjLoaders as configurables and make
jointcal.__init__() take a butler argument to configure them. JointcalRunner
has to be derived from ButlerInitializedTaskRunner to get said butler.

Clean up "default" filter handling somewhat (see DM-9093 for what else needs to
be done), to use the most common filter in the dataRefs. CollectLSSTRefStars
now takes the actual fluxField, instead of building it on the fly.

Add jointcal._do_load_refcat_and_fit() to do both photometry and astrometry
via kwargs.

Add metrics dictionary in jointcal.py, for basic values to test, that may
turn into validation/verification SQUASH tests later.

associateCatalogs now properly cleans up when run a second time. This would
be best handled via better memory management, but that's DM-4043.

Fix some range-based for loops and cleaned up pointers/references.

Remove bit-rotted Jointcal class and other unused code.

More renaming of badly-named methods and variables. Filed DM-9135 to deal
with this in bulk.

Print some interesting values in when running JointcalStatistics verbosely.

Fixed a bug I had introduced in RefStar re:setFittedStar behavior with nullptr.

Add hsc config for filterMap and tweak test to work with it.

Add every star number and chi2 metric for every test dataset.

Metrics tests always occur in JointcalTestBase._runJointcalTask(). Default
to checking every metric that gets returned, so they'll fail if that metric
isn't defined by the test method. A metric can be skipped by setting it to None.

Tweak metrics after solving common tangent plane problem.

tweak metrics for sorted dataRefs