generalize legacyhalos code to run on other samples

[moustakas]

Major cleanup and refactor of the core code to be able to accommodate other samples more easily (here, a comparison between DECaLS and HSC), with also a close eye at revisiting the legacyhalos sample.

Other pieces:

• Additional CCD-level QA for the web page.
• Custom mask and sky subtraction (written out to a new -custom-mask.fits.gz file).
• Cleaner differentiation between "custom" and "pipeline" data products.
• When ellipse-fitting, use median integration (with iterative sigma-clipping) instead of bilinear (much less noisy, but takes a lot longer).
• More parallelization.
• Lots more documentation and cleanup.

[moustakas]

Some output on a single large, central galaxy (not part of the legacyhalos/redMaPPer sample): http://portal.nersc.gov/project/cosmo/temp/ioannis/hsc-vs-decals/