Honor stderrthreshold when logtostderr is enabled

[pierluigilenoci]

What

Bump k8s.io/klog/v2 to v2.140.0 and opt out of the legacy stderr threshold behavior so that -stderrthreshold is respected even when -logtostderr=true (the klog v2 default).

Why

klog v2 defaults -logtostderr to true, which silently ignores -stderrthreshold. All log levels are unconditionally sent to stderr, making severity-based filtering impossible for log aggregation systems.

This was reported in kubernetes/klog#212 and fixed upstream in kubernetes/klog#432.

How

• Call klog.InitFlags(nil) in main() to register klog flags on the standard flag set
• Set legacy_stderr_threshold_behavior=false to opt into the corrected behavior
• Set stderrthreshold=INFO as default to preserve current behavior (all logs still go to stderr)
• Users can now override via -stderrthreshold=WARNING or -stderrthreshold=ERROR

[pierluigilenoci]

Hi — friendly ping. Is this PR still on the radar for review? Happy to rebase or make changes if needed. Thanks!

[pierluigilenoci]

Hi — friendly follow-up. The CI failures appear to be infrastructure-related (the image build can't push to ghcr.io from a fork PR, and integration tests depend on that image). The code changes themselves are minimal (klog v2 dependency update). Would you be able to review when you get a chance? Thank you!

[pierluigilenoci]

Hi — friendly follow-up on this PR. The CI seems to be failing due to a permissions issue when the workflow tries to push to ghcr.io/kubenetworks/ from a fork (error: denied: installation not allowed to Write organization package). This is expected for fork-based PRs and cannot be fixed from my side.

Would it be possible to either:

1. Configure the workflow to skip the image push step for PRs from forks, or
2. Let me know if there's another way to satisfy the CI requirements?

Happy to make any code changes needed on my end. Thanks!