Alternative ASE/MLFF batch mode

[esoteric-ephemera]

Possibly supersedes #1196. Tagging @naik-aakash and @JaGeo for thoughts.

The goal here is to make smaller structural changes overall to the jobs/workflows and ensure that each calculation maps to one output document

Changes:

• Allow for submission of a list of pymatgen Molecule or Structure to AseRelaxMaker and thus ForceFieldRelaxMaker / ForceFieldStaticMaker
• When submitting a list, a list of TaskDocuments are returned
• Cache ASE calculator once loaded to permit batch mode

[JaGeo]

@esoteric-ephemera looks great to me!

Did you think about a way to benchmark if there is indeed an advantage timewise, e.g. when executing in jobflow only. I think it will simplify execution with jobflow-remote!

[esoteric-ephemera]

I can try to throw together a quick benchmark. Any concerns about the change in output documents? Not sure how much you've been using #1196 to do calculations, but this would change how the outputs are structured (here: a list of TaskDocument vs. 1196: TaskDocument having a list of structures/energies/etc.)

[JaGeo]

We haven't used it much since the PR. There were some plans initially but we stopped this a while back. As long as the phonon workflow runs, I am happy.

[JaGeo]

My hope is that we might build a thermal conductivity workflow in the future. Here, the batch mode might be really useful as well.