In working on this, I noticed some cleanup that needs to be done in the use of forced vs. unforced catalogs for various plot (revealed when I became aware that the detect_isPrimary flag does not get persisted in the forced catalogs...neither do the calib_psfUsed & calib_psfCandidate flags).

Increasing the scope to make accommodations and include these cleanups.

The detect_isPrimary flag is actually too strict for our per-tract analysis, so I have accommodated just selecting on detect_isPatchInner and detect_isTractInner separately (in addition to deblend_nChild = 0 and merge_peak_sky = False).

Seeing as you were also recently reminded of the importance of these flags, would you mind giving this a look.  I have run a few examples for coaddAnalysis and colorAnalysis based on Hsin-Fang Chiang's w_2018_22 RC2 reprocessing, so the two sets of plots can be directly compared.  As expected, the total number of sources has decreased in the rerun based on this branch, but the statistics and color locus fits are largely unchanged.  Links to mine and Hsin-Fang Chiang's plots are, e.g.:

Coadds:

https://lsst-web.ncsa.illinois.edu/~lauren/lauren/DM-14874/w_2018_22/WIDE_GAMA15H_9615/plots/HSC-I/tract-9615/

https://lsst-web.ncsa.illinois.edu/~hchiang2/RC2_w_2018_22/pipe_analysis/qaPlots/HSC-I/tract-9615/

Colors:

https://lsst-web.ncsa.illinois.edu/~lauren/lauren/DM-14874/w_2018_22/plots/color/tract-9615/

https://lsst-web.ncsa.illinois.edu/~hchiang2/RC2_w_2018_22/pipe_analysis/qaPlots/color/tract-9615/

Note that I changed the title of this ticket so as not to be misleading.

Looks good to me! (modulo those accidental extra characters at the top of that script file)

Thanks, Tim!  Superfluous characters removed and merged to master.