[WIP] Refactor QHACalc and PhononCalc

[rul048]

Summary

Major changes:

• fix 1: QHACalc now works with relax_structure=False and ASE Atoms input (fixes [Bug]: QHACalc does not work with relax_structure=False and an Atoms object as input #152). Added to_pmg_structure() conversion before lattice scaling to ensure the structure always has apply_strain.
• fix 2: Output temperatures array now has the same length as all other property arrays (fixes [Bug]: The QHACalc temperatures and Gibbs free energies are not the same length #150). Previously temperatures had one extra element because PhonopyQHA internally clips output arrays for finite-difference derivatives.
• fix 3: Clarified np.transpose() usage on thermal property lists by inlining the QHA construction, making the data shape self-evident from context (fixes Unclear why np.transpose is being called on a list #130, supersedes Fix: Clarify np.transpose usage in QHA calculation (Issue #130) #146).
• refactor: Simplified QHACalc by removing unnecessary helper methods. Core logic is now inline in calc() for readability.
• refactor: Simplified __init__ write parameter resolution and _write_output_files with data-driven loops.
• ci: Added setuptools to ci and dev dependencies to fix maml test failures (missing pkg_resources in Python 3.12+).

Todos

• Tests: Reference values in test_qha.py may need updating if the matgl model version differs from the one used to generate the original expected values.

Checklist

• Google format doc strings added. Check with ruff.
• Type annotations included. Check with mypy.
• Tests added for new features/fixes.
• If applicable, new classes/functions/modules have duecredit @due.dcite decorators to reference relevant papers by DOI.

[Andrew-S-Rosen]

Hard coding write_phonon as False seems like it may not be ideal. What if the user wants to store the phonon properties (i.e. the YAML) out for each run to inspect the results later? Currently, there is no mechanism to do so. Of course, even if one were able to set it to True, we'd need to make sure there is no overwriting with each scale factor, so the scale factor would need to be appended to the filename or the files stored in a unique subdirectory to prevent overwriting. Also, if someone passes phonon_calc_kwargs={"write_phonon": True}, they will currently get an error.

I recognize the original code has write_phonon=False, so we can make that a separate issue report and address later if preferred, but since the I/O may influence the refactoring, I bring it up here.

[Andrew-S-Rosen]

A similar comment may hold for the phonon_calc_kwargs, since specifying any of the file I/O flags to True will cause each scale factor to overwrite the files of the prior scale factor run I think.

[Andrew-S-Rosen]

There are many changes here (like to the test suite) that seem superfluous. Given the presence of several bugs in the original QHACalc, I remain a bit hesitant about the large number of cosmetic changes proposed here. Note that the proposed refactor conflicts with some of my important bug fix PRs, like #159.

Would it be possible to have the PR just fix #150 and #152?

[rul048]

Thank you for the feedback. I will take your suggestion into careful consideration. I plan to first review and merge your bugfix PRs and then revisit the refactoring to make sure it does not conflict with those changes.

[Andrew-S-Rosen]

Alright, I am done with my major PRs for QHACalc with #163, #168, and #174. Thank you and sorry that it may have caused many conflicts with your PR here. I appreciate your help!