Implement MRAC rescaling

We were hoping to have this handled by pet-rd-tools, but it is harder than anticipated, see https://github.com/UCL/pet-rd-tools/issues/46

Given that it depends on how STIR exactly handles DICOM, Interfile etc, it seems safest to do it in STIR. In any case, we need to know how much to post-filter the MRAC.

Best would be to implement this as a DataProcessor: take (normally DICOM) image in, find originating_system, find scale factor, post-filter and natural sampling, chain-up data-processors accordingly (although admittedly this would need a new data-processor that calls zoom_image)

Obviously this all applies to CTAC as well. For that we have HUToMuImageProcessor.

Of course, this still doesn't put things in the right place until we merge #181 and #618 (and know about gantry-offsets). We have to rely on the usual strategy of NAC-recon followed by registration (as in https://github.com/UCL/STIR/blob/master/examples/GE-Signa-PETMR/register_GEAC.sh)

Start by reading the DataProcessor design and src/include/stir/HUToMuImageProcessor.h, then inspect examples/GE-Signa-PETMR/register_GEAC.sh. Determine how DICOM and Interfile expose originating_system, scaling, post-filtering, and natural sampling, including the proposed zoom_image processor chain. Done means MRAC and CTAC can use the processor safely, with behavior aligned to issues #181 and #618.