Support apptainer instead of singularity

[ffelten]

• singularity -> apptainer Support Apptainer #148
• Use advised env variables: https://scicomp.ethz.ch/wiki/Apptainer#Settings might solve Slurm scripts caching #140
• I also removed the tmp/ mount logic according to what @fgvangessel-umd said and the above point. Frank's feedback on using Slurm #151

I'm not sure how to test this though?

[g-braeunlich]

@ffelten In hpc-zaratan there are some conflicting changes with your branch.
Could we merge his branch first?

Instead of the changes of the commit Support apptainer instead of singularity, I would propose

class Apptainer(Singularity):
    """Apptainer."""

    name = "apptainer"
    executable = "apptainer"

So we support both apptainer and singularity.
Already have this in my working tree.

[g-braeunlich]

@ffelten May I cherry-pick 6a8339e to this branch, replacing Use apptainer instead of singularity?

[ffelten]

@g-braeunlich sure! Good idea

[ffelten]

Note: my student just confirmed that this new version does not have clashes relating to MPI (from different jobs) writing in the same tmp folder.

[g-braeunlich]

@ffelten I cherry-picked and rewrote the history a bit (squashed your commits, changed the commit message) into a new branch: support_apptainer-rebased.
To view effective changes:

git diff origin/feat/support_apptainer origin/feat/support_apptainer-rebased

[g-braeunlich]

After my changes, my only suggestion would be to use tempfile.TemporaryDirectory instead of /tmp (increase portability)

[ffelten]

I guess we should close this one?

[g-braeunlich]

The new branch was actually meant as a proposal to replace this branch, e.g. locally, on feat/support_apptainer:

git fetch
git reset --hard origin/feat/support_apptainer-rebased
git push -f

[ffelten]

@g-braeunlich done!

Actually, I'm not sure we want to keep supporting singularity; Euler and Zaratan support Apptainer. Moreover, the last commit on the repo was 3 years ago.

[g-braeunlich]

OK. I reverted to apptainer only

[g-braeunlich]

After my changes, my only suggestion would be to use tempfile.TemporaryDirectory instead of /tmp (increase portability)

And did you see this?

[ffelten]

And did you see this?

Missed it. We can use it indeed. I see you're on the branch. Do you want me to do it or you do it?

[g-braeunlich]

OK. I ll change that

[g-braeunlich]

Changed. I not went for tempfile.gettempdir(), as it does not contain an unpredictable directory and does not include directory cleanup.

[g-braeunlich]

Rebased on main (after the multi os PR)