Feature request: asymmetric errors

[ossk]

For a number of observational data we have asymmetric error bars. This cannot be adequately fitted at the moment. Unfortunately, this is not foreseen in the underlying astropy NDUncertainty so that it would require a major code change where basically two NDUncertainty arrays are swept through and separately used in plotting and fitting if both are given.

[mpound]

I can see the usefulness of this. Might be tricky to implement.