build(nvct): Publish and consume task-action fix in self-hosted stack

Problem

The source fix from #1042 is merged, but the current self-hosted stack still deploys nvct-service-oss:1.5.9-hotfix.1, which predates that change. A fresh local multi-cluster BDD run therefore creates an ICMSRequest with spec.action: RequestICMSInstances for an NVCT task. NVCA treats it as a function request, creates zero instances, and the task exceeds its queue wait.

This blocks the NVCT scenario in TestMultiClusterHelmfile even though control-plane install, compute registration, operator rollout, and backend health all pass.

Scope

• Publish a self-hosted NVCT image containing #1042.
• Update the NVCT chart or stack version consumed by the self-hosted Helmfile.
• Verify the task request reaches NVCA as RequestICMSInstancesForTask.
• Rerun the local multi-cluster NVCT task smoke.

Acceptance criteria

• A fresh self-hosted install no longer emits RequestICMSInstances for NVCT task launches.
• TestMultiClusterHelmfile completes its @nvct-task-api scenario.
• The published image supports the architectures required by local and EKS validation.

References

Follow-up to #1032 and #1042

Locate the self-hosted Helmfile, NVCT chart or stack image-version references, and the TestMultiClusterHelmfile entry point for @nvct-task-api. Start by checking which image currently supplies nvct-service-oss and how its architectures are published. Done means a fresh install sends RequestICMSInstancesForTask, the NVCT scenario passes, and the image supports local and EKS validation architectures.