Investigate and fix jobs with empty SpackJobData values

[jjnesbitt]

• It seems there always been some portion of jobs with "empty" values
• Up until around May 29 2025, the empty values had their respective job_type field correctly filled out, with the job_size and stack values being set to "1"
• After May 29 2025, when a job is missing these values, they are simply empty
• This does not seem to simply be a case of UO runners. Often times these jobs are run in the cluster.
• Often times these jobs are very quick (under 10s), which means that it's possible prometheus misses the job entirely. However, there are jobs with longer runtimes (tens of minutes) that also exhibit this problem.
• If these are non-build, non-generate jobs like no-specs-to-rebuild or rebuild-index, there won't be any artifacts associated at all. So those files can't be used to retrieve this info
• It seems that in cases where a build job has these values missing, the package_id value is also missing.
• For UO runners, if there are no available artifacts (non-build/non-generate job, job timed out, etc.), there is no other way for us to retrieve these values.

We should try to make the following improvements:

1. For non-cluster jobs, we should ensure there's always a way to retrieve these values. At the moment, there's many legitimate cases where we have no way to retrieve these values (non-build / non-generate jobs).
2. Investigate and fix any flakiness for retrieving values from prometheus for jobs who have some minimum runtime (e.g. 10 minutes, just to rule out the job being too short to retrieve values)
3. Find an alternative way to retrieve these values when jobs are too short to retrieve from prometheus, and will not produce any artifacts. This is essentially the same case as point 1.

Here is a metabase question that will show recent jobs that exhibit this issue.

[jjnesbitt]

It seems that the stack at least could be retrieved by retrieving the trigger job of the parent pipeline, that is apparently named {stack}-build.

[jjnesbitt]

To retrieve other information about a job (like the compiler, compiler version, etc.) the following can be done:

1. Start with failed job
2. Get job's pipeline
3. Get the parent pipeline of that pipeline
4. In that parent pipeline, there will be a generate job of the same name as the trigger job for the child pipeline (in this case, ml-darwin-aarch64-mps-build and ml-darwin-aarch64-mps-generate)
5. That job will have a spack.lock file in its artifacts.