Add Q-resolution support

[SimonHeybrock]

This is not ready for full review / merging, just gathering Scientist feedback

Fixes #111.

To try this out use, e.g., sans2d.ipynb and add:

workflow[sans.qresolution.SourceApertureRadius] = sc.scalar(0.020, unit='m')
workflow[sans.qresolution.SampleApertureRadius] = sc.scalar(0.008, unit='m')
workflow[sans.qresolution.CollimationLength] = sc.scalar(4.0, unit='m')
workflow[sans.qresolution.DeltaR] = sc.scalar(0.008, unit='m')
workflow[sans.qresolution.ModeratorTimeSpreadFilename] = ('ModeratorStdDev_TS2_SANS_LETexptl_07Aug2015.txt')
resolution = workflow.compute(sans.qresolution.QResolution)

Then play with resolution.bins.mean() or resolution.bins.max(). Note that if WavelengthBands is configured, the resolution is computed for each band, as one would expect.

Questions to science reviewer:

1. Is the math correct? I have used https://docs.mantidproject.org/nightly/algorithms/TOFSANSResolutionByPixel-v1 and the linked Mantid code as a reference. Note that the docs there claim some small-angle approximation was made. I guess that is quite wrong for the Loki window-frame banks? What should we do about this?
2. What do you require as final output? Right now I am returning binned data (see docstring), which may be useful for analysis, but I presume ultimately you need something else?