EAMxx: fix ghci-snl-cpu standalone machine setting

[bartgol]

We were missing the gator initial mb env var, which caused the valgrind build to run out of memory.

[github-actions]

PR Preview Action v1.4.8
🚀 Deployed preview to https://E3SM-Project.github.io/E3SM/pr-preview/pr-6774/
on branch gh-pages at 2024-11-23 02:37 UTC

[bartgol]

@jgfouca I manually triggered the memcheck nightlies from this branch, and it does appear to fix things. See here for the log, or cdash for the fresh report.

The valg build still has the issue of generating an empty suppressions file, but maybe we can work around that. E.g., we could store a supp file inside the container or in the repo, and call it a day. However, in the midst of MPI false positives, I think I saw some potentially real errors coming from IOP, p3, and maybe Homme. I just skimmed, so I am not 100% sure, and I have to call it a day (a week actually), but I will continue on Monday.

I do think we should merge this PR though, so we start to get the valg build to report to cdash.

[rljacob]

Seeing this error in the CI:

[ci (SMS_D_Ln5_P4.ne4pg2_oQU480.F2010-SCREAMv1-MPASSI.ghci-oci_gnu)](https://github.com/E3SM-Project/E3SM/actions/runs/11983339941/job/33428932378#step:5:1)
Value cannot be null. (Parameter 'ContainerId')

[mahf708]

Seeing this error in the CI:

[ci (SMS_D_Ln5_P4.ne4pg2_oQU480.F2010-SCREAMv1-MPASSI.ghci-oci_gnu)](https://github.com/E3SM-Project/E3SM/actions/runs/11983339941/job/33428932378#step:5:1)
Value cannot be null. (Parameter 'ContainerId')

"too many requests" error ... I was hoping it would go away. Let me take a look. (At some point, we will figure out a way to cache all the container nearby so that we don't need to pull it like we do now...)

[bartgol]

The only fail on CUDA is the known issue with mam4 test (likely a non-determinism), which is orthogonal to this PR. Merging.