Broadcast uncertainties

[jl-wynen]

          > Does the result from broadcast_uncertainties work with scipp.loookup?

It should. We use it here: https://github.com/scipp/essdiffraction/blob/149203c29be6dd6e5dc49bf910015b308c0afa9e/src/ess/powder/correction.py#L56-L59

The monitor uncertainties are necessarily large at the minimal and maximal wavelengths transported by the guide, where the incident intensity goes to zero -- but the uncertainty relative to the detector uncertainty is probably always small due to the small cross section for inelastic scattering.

I think this deserves some more investigation once we have the full workflow. We should ensure that we can ignore correlations around the edges of the data. Or avoid using those data.

Originally posted by @jl-wynen in #24 (comment)