[Automated] Refresh kagent UI screenshots (live cluster)

[github-actions]

Automated refresh of the live-cluster kagent UI screenshots.

A diff here means the kagent UI changed. Review the images before merging, and
check whether the prose in the guides that embed them is now wrong too.

Captured against a kind cluster running the chart versions the install guide pins,
so these images show what a reader who follows the docs actually gets.

Relative timestamps ("2 minutes ago") are left visible rather than masked, so a
small diff confined to those is expected and not a UI change.

Generated by the Refresh kagent UI screenshots workflow.

[Rachael-Graham]

Closing unmerged: this diff is per-run noise, not a UI change.

The Substrate page reads v0.2.0-beta5 in both the committed baseline and this capture, and the layout is identical. The only differences are values that change on every run:

• worker pod name suffix (kagent-default-774d4496bd-4snsc → -hz422)
• pod IP (10.244.0.24 → 10.244.0.23)
• relative timestamp (read just now → read 1.0s ago)

The agents capture differs only in text metrics, from the baseline having first been captured off an ubuntu runner.

Because those values are regenerated by every kind install, the cluster job will open an identical PR on every run. The PR body already excuses relative timestamps; the pod name and IP are not covered. Tracking that as a follow-up — the regions need masking in launch-ui.spec.ts before this job keeps running on push.

The real UI change from this run is in #523.