Color Matrix Alpha is not handled correctly

[LemonInTheDark]

In BYOND, the raw alpha channel in color matrixes (a -> a) is treated as partially linked to the alpha var. Updating the aa channel will mirror alpha to match, and updating alpha will set the aa channel to 1.

In opendream they appear to be treated as two different, separate things. In addition, it seems as if the aa channel straight up does not DO anything. Other components of the color matrix are allowed to muck with the alpha of the world, but aa is seemingly totally ignored.

[wixoaGit]

We currently store matrix and non-matrix color as two values in different fields, and do some logic to decide which to use. There's a good chance this logic is broken.

A fix for this would preferably only store the color matrix on appearances and convert between the two when DM is using atom.color.