Paul, up for an LSST-side review?

This is a plugin interface (in the new meas_base framework) for my new shapelet-PSF-approximation code. The deep algorithmic code this delegates to has already been reviewed, so this is just the Python layer that lets us run it as a plugin. Most changes are in meas_multifit, on branch u/jbosch/DM-1223. There are also some small changes in meas_base (u/jbosch/DM-1223) that aren't really directly related, except that they were problems I had to fix in order to get the new plugins (and particularly their tests) working in meas_multifit. Reviewing commit-by-commit will probably be easier than parsing the full diff for both packages.

If you want to build these branches yourself, you'll need u/jbosch/DM-1224 of shapelet, and a moderately recent master of everything else.

meas_base review done at https://github.com/LSST/meas_base/pull/4

meas_multifit review done at https://github.com/LSST/meas_multifit/pull/4

Nothing substantial. I do think you might have done the multiple inheritance using super and some introspection, but I think the case you've identified (different arguments for different subclasses) is not well handled in the language in general and you appear to have been careful to avoid the gotchas (e.g., no diamond inheritance, not using the same variable names), so I won't object to your solution.