Change the NearestNeighbours code to work on spectra rather than pixels

[Martyn Gigg]

An instrument can have detectors grouped at the hardware level, i.e WISH and MERLIN, so the nearest neighbours code needs to change to find nearest spectra rather then nearest detectors.

[Martyn Gigg]

(In [11920]) Added an interface for the spectra-detector map and updated it with iterator capabilities. Also added a width member to BoundingBox. Refs #3052

[Martyn Gigg]

(In [11925]) Consolidate some code (well it will be when I commit the bit that uses it later). Re #3052

[Martyn Gigg]

(In [11949]) Re #3052. Nearly final touches to changes for nearest neighbours grouping spectra not detectors. It still returns det_ids which isn't ideal but the core functionality is doing what is should do so this is essentially a checkpoint.

[Martyn Gigg]

(In [11965]) The nearest neighbours method now on the detector class now returns a mapping of spectrum number to distance rather than detector ID. Re #3052

[Martyn Gigg]

(In [11966]) Fix PythonAPI build Re #3052

[Martyn Gigg]

(In [11974]) Change the boost graph in the nearest neighbours class to store the distances in real space not our scaled version. Re #3052

[Martyn Gigg]

(In [11976]) Fix for crash in release mode where an array went out of bounds. Re #3052

[Stuart Campbell]

This ticket has been transferred to github issue ​3899