Support full output in curve_fit

[jl-wynen]

scipy.optimize.curve_fit has an argument full_output which makes it return additional objects.

At the moment, our wrapper raises ValueError: too many values to unpack (expected 2) with this argument.

We should catch this case and forward the extra return values. I'm not entirely sure how to handle them because they contain numpy arrays. In particular a Jacobian matrix which cannot be represented as a scipp array in a clean way.