Add ability to send positions to CP2K via file

Introduces the ability to side-step issues in passing large atomic structures to CP2K through stdin for MPI applications. Adds a set_pos_file option that sends positions via file rather than stdin, which is a route others have used successfully and one I've confirmed fixes issues with ASE.

My PR to CP2K which supports the new route has been merged, but we'd need to test against the main branch until the next CP2K release occurs.

Checklist

• I am familiar with ASE's contribution guidelines.
• Doc strings in code changed in this MR are up to date.
• Unit tests have been added for new or changed code.
• Issue is resolved via "closes #XXXX" if applicable.
• Add warning about using the "set_pos_file" for structures larger than a few dozen atoms
• Check the shell version so we can warn users if their CP2K is out of date.

changed title from Draft: Attempt a set_pos_file option to Draft: Add ability to send positions to CP2K via file

changed the description

added 1 commit

• 615b8de0 - Do not delete file, that - somehow - leads to CP2K stalling

Compare with previous version

added 1 commit

• d212b801 - Added warnings about versions, original error

Compare with previous version

marked the checklist item Add warning about using the "set_pos_file" for structures larger than a few dozen atoms as completed

marked the checklist item Check the shell version so we can warn users if their CP2K is out of date. as completed

marked this merge request as ready

I'll push a unit test soon, but it won't work. Shall I mark it an xfail so that it'll eventually pass when CP2K is updated?

changed the description

mentioned in issue #939

mentioned in issue #1477

added 1 commit

• 4a108514 - Add unit test, instructions for after CP2K release

Compare with previous version

marked the checklist item Unit tests have been added for new or changed code. as completed

marked the checklist item Issue is resolved via "closes #XXXX" if applicable. as completed

Any reason why we are printing to the file instead of just writing to it directly via fp.write()?

Apologies for the delay, @WardLT. Just one small question below before I go ahead and merge.

No, I don't have a reason beyond "habit." Would write be more performant? Print flushing the write buffer does seem unnecessary

I can change that. Won't be for a few days, until I'm back working off the computer which has the newest cp2k

Mostly, I was just genuinely curious. It really shouldn't make much of a practical difference here. I haven't actually seen someone do this before in the wild! No need to revisit this MR. I'll merge.

merged