Maskbits II

I talked to @moustakas with respect to #608 and the suggestion is to separate populating maskbits from within select mock targets and instead pre-burn maskbits files for the mock targets by rescoping the select_randoms script to read in a set of mock (ra, dec) and continue from those, rather than from a set of Poisson points, when flagged appropriately.

Few points:
-- We're going to have 1000s of mocks to have to do a mask calculation for, for each tracer, so speed
is important.
-- We'll also have the use case where we want to save depths, etc. rather, than just an in/out flag for
a given tracer. Note that the in/out decision will still require NOBS etc, i.e. it'll call is_in_elg_mask.
-- It might be better to simply 'sphere match' to the random catalogue? I've heard arguments both ways.

There's a cosmosim call on Thursday likely addressing some of these issues. Putting this out there before I try anything again.

Start with issue #608 and the select_randoms script, then trace how it currently begins from Poisson points and calls is_in_elg_mask. Resolve the proposed mock (RA, Dec) input, maskbits/depth outputs, and sphere-matching approach after the mentioned cosmosim discussion; the issue does not yet define a concrete done state.

Written by the indexing model from the issue text.