More robust grouping by angle

[jl-wynen]

The BIFROST workflow groups by sample and detector angle a3 and a4. The implementation needs to be extended for:

• Varying a4.
• Intermediate angle values that are recorded while the sample or detector moves. Those time regions likely need to be discarded.

The implementation also needs to be benchmarked for long measurements with many log entries for the angles. We may be better off, avoiding sc.lookup in that case and determine bin edges first and then bin by those edges.

[g5t]

Another pitfall that I've now encountered is that some detectors may record no events for some angles, then the lookup-method 'misses' measurement periods with null observations.

• These missing observation periods can lead to incorrect values reported when later collecting intensities for slices and cuts.
• More immediately they will prevent detector merging if any one detector 'misses' a different set of angles than the others.