Support variable-TR, variable-flip-angle protocols in MCR-MWI processing and fitting

[samuelmelke]

Adds support for variable-TR, variable-flip-angle protocols in the MCR-MWI
preprocessing and fitting stages. Constant-TR behaviour is unchanged.

+qb/+workers/VFAprepWorker.m — make_syntheticT1_M0 assumed one TR for
all flip angles: DESPOT1 was fitted with whichever TR was read last, and every
synthetic T1w target was built with that same TR. It now reads
RepetitionTime per flip alongside FlipAngle and dispatches on whether the
TRs differ: despot1_mapping for constant TR, despot1_VTR for variable TR.
GRESignal takes the per-flip TR, which is identical to the previous
behaviour when the TRs are equal.

+qb/+workers/MCR_GPUWorker.m — detects a variable protocol (differing
TR, echo counts, or TE grids across flips) and dispatches to
gpuMCRMWI_VFAVTR instead of gpuMCRMWI, concatenating echoes along dim 4
with a te_indexrange mapping so unequal echo trains per flip are handled.

.gitmodules / dependencies/vTR-qMRI — adds vTR-qMRI as a submodule,
providing despot1_VTR and gpuMCRMWI_VFAVTR.

[marcelzwiers]

Hi Samuel, can you do your pull request against dev instead of main?

[samuelmelke]

Done.

[marcelzwiers]

Question, did you manually edit the .gitmodules file or did you use:

$ git submodule add -b main https://github.com/samuelmelke/vTR-qMRI.git dependencies/vTR-qMRI

I'm asking because I had created a lot of issues myself in the past when I edited the file manually...