generate tilepix.fits and tilepix.json files from reduced data products

[moustakas]

Addresses #1374 by reading the FIBERMAP tables contained in the healpix-organized spectra- files rather than in the preproc directory, thereby guaranteeing that the tilepix.fits and tilepix.json files will only contain healpixels which have fully reduced spectra on-disk (for a given data release).

I've run a few small tests and verified that it addresses the examples given in #1374. However, I'm running a full-Everest test right now with

desi_map_tilepix --reduxdir /global/cfs/cdirs/desi/spectro/redux/everest --nside 64 -o new-tilepix.fits

and will report back when that is complete.

Note: I'm not sure where/when desi_map_tilepix is currently invoked in the pipeline processing steps, but if this PR is accepted then it means that it will have to be run at the end of a data release after the healpix directory has been fully populated.

[coveralls]

Coverage remained the same at 25.201% when pulling f202c1e on fix-tilepix-map into ea76485 on master.

[moustakas]

It's possible this PR also addresses #1379, although I'm a little unclear about this statement from @sbailey because this PR only affects the tilepix.json and tilepix.fits files, not the tiles-{specprod}.fits file:

These [dither] test tiles shouldn't be in tilepix.json, and tilepix.json and tiles-{specprod}.fits should have the same tiles, and all of those should match what is actually on disk in healpix/{survey}/{faprgrm}/

It's possible (quite likely!) that I'm just muddying the waters here and, if so, apologies.

[moustakas]

Also fixes #1373.

[sbailey]

Thanks. I expect we'll make a few more last-minute changes to the healpix workflow after Fuji is underway, but healpix has to wait for all tile-based processing to be finished first so it doesn't have to be a blocking factor. I'll merge now to get it in the first Fuji tag, and if we need to make more updates later ok.