B1 mapping improvements

[lukeje]

Merge the implementation of (S)DAM B1 mapping from @tleutritz-cbs into the toolbox. Also includes a bugfix for a bug that can occur if filename lengths of the B1 input files are not equal.

[lukeje]

@kerrinpine: I've made the change that AFI (and DAM) no longer uses "abs" to get rid of the imaginary component, but rather uses "real" instead. In cases where AFI or DAM images are input in the wrong order, this now results in a large hole in the centre of the B1 map, rather than "reasonable" looking incorrect values. Hopefully this obvious error will cause people to check their input data, rather than allowing the erroneous B1 maps to propagate down the processing chain.

Unfortunately I could not find a simple robust criterion to warn if too many complex results are returned, as background voxels are expected to show a significant number of erroneous values, and the number of background voxels in a volume can be very large.

I will include these changes in our Leipzig toolbox as soon as this pull request is finalised here.

[lukeje]

Unfortunately I could not find a simple robust criterion to warn if too many complex results are returned, as background voxels are expected to show a significant number of erroneous values, and the number of background voxels in a volume can be very large.

Realised that comparing the result between the two possible orderings works for AFI (but not DAM).

[lukeje]

TODOs:

• add smoothing option to pre-processed B1 (request from @mfcallaghan)
• more testing that changing 3D-EPI-SE/STE algorithm does not cause significant changes (request from @antoinelutti)
• document expected order of 3D-EPI-SE/STE input if input-checking is disabled (request from @nbeliy)
• add switch to read 3D-EPI-SE/STE flip angles from json metadata (request from @nbeliy)