mrcat: Change positional argument structure

[Lestropie]

Currently, mrcat enforces at the point of command-line parsing that there are at least two input images. It does this by having Argument image, then Argument image2 with .allow_multiple().

I have two issues with this design:

1. Consider Implement Pydra code-generators #2665. Ideally, the utilisation of mrcat in a Pydra workflow would look something like:

       images = ['image1.mif', 'image2.mif']
       workflow.add(mrcat(inputs=images, ...))
   

   However with an automatically-generated Pydra interface, what would actually be required
   would end up being a rather clunky:

       workflow.add(mrcat(image1=images[0], image2=images[1:], ...))
   

2. This mechanism is absent from other commands where it is equally applicable; eg. mraverageheader, transformcompose. So the software is not internally consistent in this regard.

For the first point in particular, I think it makes sense to have a single positional input argument with .allow_multiple(). But I am open to a counter-argument.

---

Hypothetically, we could introduce some other Argument modifier that would set the minimum number of occurrences. This would allow catching command mis-use at the point of command-line parsing. But we would then need to consider how to reflect this information in the various __print_usage__ interface exports. Any thoughts?

[github-actions]

clang-tidy review says "All clean, LGTM! 👍"

[Lestropie]

After further consideration in the context of Pydra workflow construction, it's possible that changing the mrcat operation with a single input from an Exception to a WARN() may be preferable.

[jdtournier]

Fully agree with these changes. Not sure why we'd implemented it that way in the first place, in hindsight...

[jdtournier]

After further consideration in the context of Pydra workflow construction, it's possible that changing the mrcat operation with a single input from an Exception to a WARN() may be preferable.

Personally, I don't think this should even raise a warning. There's nothing theoretically wrong with passing a single input image to mrcat, and I can see situations where some script may perfectly legitimately need to concatenate one or more inputs, and raising a warning every time the user happens to provide a single input would just cause needless confusion. Just my 2¢...

[github-actions]

clang-tidy review says "All clean, LGTM! 👍"