Complete the baseline summation integrator

[jbeilstenedmands]

Complete the baseline c++ integrator implementing summation integration.

Input can be the output from the baseline indexer, predicted data, or predicted data containing bboxes.

Outputs a reflection table with sufficient content to be usable for dials.symmetry and dials.scale.

Slow as not yet parallelized over cpu threads, but useful for testing and comparisons. Close to dials equivalence, not expected to be fully equivalent as slightly different approach used for classifying foreground and background.

[jbeilstenedmands]

Example differences compared to dials summation on 100 images of thaumatin from i03.
This was with dials predictions and bboxes and fixed sigma parameters, so just assessing the differences in the foreground/background classification, and the overall differences seem minimal.

[jbeilstenedmands]

Ok, there is some difference coming from the slightly different foreground/background segregation algorithm compared to dials. However this baseline integrator is currently missing robust background modelling - dials removes outliers from the background array, even for constant3d model (using glm method), so this needs adding before we can check equivalence.