Add diagnostic diff reporting in regression tests + remove ifort array temporary warning

[BenjaminTJohnson]

Summary

This branch extends regression tests to report min/max/relative difference statistics across key CRTM structures, improving diagnostics when comparisons fail. It also removes Intel Fortran debug warnings about array
temporaries created when allocating FitCoeff objects.

Changes vs develop

• Added shared diagnostics include with diff-stat reporters for CRTM structures.
• Wired the new diagnostics into forward/adjoint/k_matrix/tangent_linear regression test drivers (e.g., ClearSky, Simple, SOI, AOD, SSU, Zeeman, etc.).
• Avoided ifort array temporary creation in FitCoeff_SetValue by storing SHAPE(C) in a local dimensions array before calling FitCoeff_Create.

Testing

• Manual debug build and run; ctest warnings no longer appear (for FitCoeff)
• Whenever there is a difference in results (e.g., DEBUG vs. RELEASE) in the ctests comparisons, you should see additional information about differences. I'll probably expand upon this a bit later, but for now it gives some additional context rather than just saying "results are different".

[chengdang]

All ctest passed.
Just to make sure I understand it correctly, this bug is raised when optical depth is very large. By "clamping" the exp (-OD), we can constrain this issue. Is this correct?

Is Compare_Diagnostics.inc a brand new module for CRTM? This is super thorough and useful!

[chengdang]

Hi Ben, I'm trying to understand this set up. With such if statement, if the total AOD is too big (is_clamped), then the total_opt_AD is not updated?

[BenjaminTJohnson]

@chengdang this PR is partially impacted by the other big one #277, so after we merge 277, I'll come back to this one. Switching to draft. I'll also review your questions at that time. Thanks.