De-reddening

[dfm]

As discussed over in #5: there is some question about whether it would make more sense to build the noise model in de-reddened color rather than observed. I think that it probably does and it might make a real difference.

@adrn: The fits file that you shared with the RVS sample from EDR3 doesn't have an excess color column. Do you know what it would take to apply a de-reddening estimate to that sample?

[adrn]

As usual, it's a long story 😅. As a first pass, if you pip install git+https://github.com/adrn/pyia and load the EDR3 data file with pyia:

>>> from pyia import GaiaData
>>> g = GaiaData('/path/to/thefile.fits')

You can get the extinction-corrected magnitudes with

>>> G, BP, RP = g.get_G0(), g.get_BP0(), g.get_RP0()

But this is only a first pass because: (1) The Gaia bandpasses are broad enough that the extinction factor needs to be computed iteratively. I'm using an algorithm that was suggested in one of the DR2 papers, but I'll need to check to see whether there are better ideas for EDR3. (2) The default uses the SFD dustmap, but many of these stars will be nearby enough that we might need to use a 3D dustmap instead. The problem there is that the most precise 3D dustmaps are only in the pan-starrs footprint, so not all-sky. I'll think a bit on this too...