fix: upgrade Kubernetes client to v31.0.0 to match cluster version

[bodom0015]

Problem

"COMPLETED" events are being skipped due to an error:

2025-01-29 18:21:02,090 [services.kubejob_service] ERROR    Unknown exception - KubeWatcher reconnecting to Kube API: '>' not supported between instances of 'NoneType' and 'int'

Approach

• fix: upgrade k8s client to v31.0.0 to match our Kubernetes version on the cluster: 1.31
• fix: print stacktrace in the logs for Unknown exceptions in KubeWatcher
• fix: adjust failure/completed case checks to avoid the NoneType comparison error above

How to Test

Deployed to mmli-backend-staging

1. Navigate to https://clean.frontend.staging.mmli1.ncsa.illinois.edu
2. Submit a new job:

>1234
asdf

3. When job completes, you should see results
   • Example: https://clean.frontend.staging.mmli1.ncsa.illinois.edu/results/209ec5fc71324e5d984dd65f8ddaa01d/1

[ckouder]

Thank you Sara! Tested with https://clean.frontend.staging.mmli1.ncsa.illinois.edu/results/880098bab87144d4b2f1f215e23504b0/1