Improve spectral model uncertainty handling #885
This PR refactors the uncertainty handling in
cdeil left a comment
I had a quick look and left two inline comments.
Concerning the Butterfly class -- the reason we introduced it was to split computation from plotting and eventually also to have a serialisation format and a place to put the I/O methods. So I think we might want to keep it, or at least discuss whether to use a plain Table or our own SpectrumButterfly class before removing it. Changing it to hold a Table attribute instead of sub-classing Table would probably be a good move (not here, someone, some day) that doesn't need discussion.