Replace mfiter loops with the MF parallel for versions

[marchdf]

Summary

Replace mfiter loops with the MF parallel for versions. These take care of a bunch of defaults for use, creates fused parallelfors for the GPU (nominally results in faster code), and generally simplifies the code.

I am looking for volunteers. I have annotated all the mfiter loops that can be simplified. I probably missed some and some might not be doable on a second glance. I did a couple to show how it is done and there are other examples in the code.

Annotations are // this one for mfiter loop canditates.

Also identified // needs openmp pragma? for loops not containing openmp and that maybe should.

I am using clang-format 18 to format the code, which causes the formatting check to break. But I am going to leave that alone so we don't waste resources and will fix the formatting later.

Pull request type

Please check the type of change introduced:

• Bugfix
• Feature
• Code style update (formatting, renaming)
• Refactoring (no functional changes, no api changes)
• Build related changes
• Documentation content changes
• Other (please describe):

Checklist

This PR was tested by running:

• the unit tests
  • on GPU
  • on CPU
• the regression tests
  • on GPU
  • on CPU