Make sure each point has enough q values to support resolution calculation

[pkienzle]

Data which is sparsely sampled, data with gaps, and the leading and tailing data points need q values different from the measured q values in order to calculate the resolution. Add support for this into Probe.

This will also be an issue for polychromatic instruments with energy-dependent sample scattering measured at a few widely separated angles. The different wavelengths effectively become separate reflectivity curves, and the resolution function is two-dimensional, with Δλ mixing between curves and Δθ mixing within the curve. The reduction software should preserve the grid structure for non-stitched data rather than trying to infer it on load.