Correct difference photometry for reference flux

[dmitryduev]

This PR implements corrections to the (difference) alert photometry if flux is present in the reference image.

Also fixes a bug in flux sign computation based on candidate.isdiffpos.

See https://github.com/dmitryduev/kowalski-archived/blob/master/kowalski/server.py#L1603

[dmitryduev]

@profjsb could you please accept the invite to K repo? I would appreciate a review of this PR from you!

[profjsb]

Looks good but it would be useful to also include tests for the scenarios where there is reference flux (ie. stellar host). It might also be a good idea to try out a known transient atop faint hosts and make sure that the resultant light curve is acceptable to the constituents of that science.

Would suggest adding a flag the alt data field in the SP posting for photometric points that have had the reference DC level added in (like {diff_flux: False})

[dmitryduev]

@profjsb suggested: add a new column to Photometry in SP, call it is_diff (which defaults to True? yes for ZTF alert data, but otherwise I'd leave it nullable) or something. Make SP display only the is_diff data for now. Let K then post (potentially) two sets of points: one with difference photometry and the other with reference-flux-corrected photometry (if such a correction was applied).