Implement numerical averages

It will be very interesting to benchmark numerical spherical averages in the incoherent approximation with the almost-isotropic approximation currently used in Mantid-Abins.

We expect that in most cases the approximation is fine, but it would be good to have a better understanding of the limiting/edge-cases and some way of detecting them.

*Duplicated from https://github.com/pace-neutrons/abinslib/issues/37*

Start by reviewing the duplicated abinslib issue 37 and the current almost-isotropic approximation in Mantid-Abins. Clarify the intended numerical spherical-average method, limiting and edge cases, and how they should be benchmarked or detected. Done should include agreed comparison criteria and an implementation that addresses them.

Written by the indexing model from the issue text.