platform.chartPath labs set kagent.serviceMonitor.enabled: true, so platform-test expects a kagent scrape the kagent line never serves

[teemow]

Problem

With platform.chartPath (a local meta chart directory) agentlab writes its stable-line lab values, which set kagent.serviceMonitor.enabled: true — the connectivity chart then renders the kagent controller metrics Service and ServiceMonitor. The dev-channel shape (platform.chartBranch) sets false with the comment that kagent main opens no metrics listener. Meta chart 4.0 (giantswarm/agent-platform#341, PR #350) runs the kagent line (kagent API v2), whose controller serves no /metrics, and defaults kagent.serviceMonitor.enabled to false in both charts; a chartPath lab of that chart still gets true from agentlab, so:

• Prometheus gets a kagent target that is never up, and
• agentlab platform-test fails at "Verifying Prometheus scrapes the platform itself" with prometheus is not scraping kagent — the check expects a kagent target whenever the ServiceMonitor exists (kagentControllerMonitored()).

Seen on agentlab v0.31.1 proving PR #350 from a checkout (platform.chartPath); a platform.valuesFiles overlay with kagent.serviceMonitor.enabled: false works around it.

Proposed solution

The lab values leave kagent.serviceMonitor.enabled to the chart's default (or set false) whenever the platform runs the kagent line — for chartPath as for chartBranch; from meta chart 4.0 on that is every lab (#138 makes the 4.0 topology the lab's shape). platform-test keeps expecting a kagent target only where a ServiceMonitor exists, so nothing else changes.

Acceptance criteria

• A platform.chartPath lab of a 4.x meta chart renders no kagent controller ServiceMonitor without an overlay; agentlab platform-test passes on it.
• The dev-channel and chartPath shapes agree on the kagent monitor value (one place in the template).