Major refactor of testing code to use new JointcalTestBase, pulling most of the
work out of the old testJointcal.py (which is now testJointcal_lsstSim.py, and
which now uses the base class). New testJointcal_cfht uses this base class. Also
improved plotting code to generate per-visit quiver plots.

testJointcal_lsstSim doesn't use the parseAndRun code, because the lsstSim data
is missing some necessary metadata. Can't solve that until the new metadata
class is built and the twinkles data is reprocessed. The test code still works
fine, though.

Improved some docstrings in jointcal.py. Left in the commented-block that used
the old StarSelector: I think we can remove it when we have a few more robust
source flags to describe good astrometric sources.

Jointcal.run() now returns a struct with the modified dataRefs and the old WCSs.

JointcalRunner now works correctly with parseAndRun, fixing DM-6915. Also made
it return a Struct for testing the results.

Added coaddName back to JointcalConfig (parseAndRun needs it).

Better documentation of relative and absolute error thresholds.

New and improved plotting code for interpreting results.

Pulled various blocks into short helper functions.

Began moving test files to _ variable names.

Forgot to commit updated cfht tests.