astropyconst20 and the CODATA version

With the changes with the default CODATA values, we seem to have a minor doctest failure. I am unsure if its my lack of understanding of how units and constants versions work underneath.

So within our constants.py we import the astropy constants like so:

import astropy.constants.astropyconst20 as astrocon

and we set our mass value (among others) like so.

physical_constants['mass'] = astrocon.M_sun

When we check the mass value it reports that it uses the CODATA 2018 whereas I thought that astropyconst20 would be CODATA 2014.

>>> from sunpy.sun import constants
>>> constants.get('mass')
<<class 'astropy.constants.iau2015.IAU2015'> name='Solar mass' value=1.988409870698051e+30 uncertainty=4.468805426856864e+25 unit='kg' reference='IAU 2015 Resolution B 3 + CODATA 2018'>

I am unsure if it a bug or something we screwed up inside SunPy.

[mhvk]

I think what is happening is that when one loads values from iau2015, the ones that involve codata.G get calculated on the spot. This works beautifully when one uses the astropyconst20 science state, since codata is set correctly to codata2014, but not if constants has already been loaded, since then the most recent codata.G is used.

This does not seem easy to change in an automatic fashion. But it would be trivial to just copy the relevant definitions of M_sun, M_jup and M_earth to astropyconst20 itself. This is not super-elegant, but I think it is OK, since all the packages involved are fixed.

[nabobalis]

Thanks @pllim.