Additional data model updates for DR8

[geordie666]

This PR updates a few remaining elements of the data model to run on the DR8 imaging surveys. Includes:

• Add the WISEMASK_W1 and WISEMASK_W2 bitmasks to the random catalogs.
• Deprecate BRIGHTBLOB in favor of using MASKBITS for targets.

The second update is necessary because MASKBITS should be identical for the targets and for the randoms, whereas BRIGHTBLOB would be marginally inconsistent between targets and randoms. See #491, which this PR should address.

[geordie666]

@moustakas: as I know you're about to start working on the mocks again, I wanted to push the (hopefully) final data model changes for DR8.

These will, in typical fashion, break select_mock_targets, although I think you probably just have to change occurrences of BRIGHTBLOB to MASKBITS.

If you were mocking up BRIGHTBLOB to be 2**0 for the mocks, note that you'll now have to make it 2**1, or bit-shift it as I did here.

[geordie666]

@moustakas: I also restored the meaning of the qso_selection==colorcuts option for the SV cuts. It should now be straightforward to bypass the QSO Random Forest selection when creating SV mocks, e.g.:

select_targets(infiles, survey='sv1', qso_selection='colorcuts', ..., ...)

[moustakas]

Thanks @geordie666. The targeting catalogs do not contain the (new) PSFDEPTH_W[1,2] columns -- do we want these?

[geordie666]

PSFDEPTH_G, R, Z and GALDEPTH_G, R, Z are not used for the targets. They are produced in the random catalogs, and then those randoms are used in QA, but they aren't touched in the target selection itself.

So, guessing that PSFDEPTH_W1, W2 will be the same, and with a view to not further unnecessarily expanding the target files, I'm not going to propagate PSFDEPTH_W1, W2. For now.

[geordie666]

Having said that, if anyone wants PSFDEPTH_W1, W2 it's probably Rongpu, so I'll ping him and check.

[sbailey]

Is this PR backwards compatible with running TS on DR6 and DR7, in addition to supporting the data model changes for DR8?

I'm ok with not including PSFDEPTH_* in this round, but we should revisit that in the broader picture of what should (not) be included. The original guidelines for what should be included in TS:

1. columns needed for provenance (BRICKID, TARGETID, RELEASE, ...)
2. columns used by target selection
3. columns likely needed for key downstream analyses (QA and/or science)

PSFDEPTH_* seems in the (3) category. Of course this was before the TS files ballooned, so now were revisiting what really needs to be included.

[geordie666]

Trying to condense a bunch of discussion on-and-off-list into a conclusion:

I'm confident that this PR IS backwards-compatible with DR6 and DR7 files (I'll check one last time before merging). And, I won't propagate propagate PSFDEPTH_W1, W2 for targeting, but I will add them to the sweeps in the legacypipe repo.

@moustakas: Unless you disagree I'll merge #493, #494, and #496 in an hour or two after running a few final tests.