helm: restore kataManager defaults

[Ayush-Rathor]

Description

This fixes a regression in the Helm chart defaults for kataManager, issue 2398.

In v25.10.0, the chart shipped a complete default kataManager block including:

• config
• repository
• image
• version

In v26.3.1, those defaults were removed, leaving kataManager incomplete when enabled through values. This caused the rendered ClusterPolicy.spec.kataManager to miss required image metadata and broke reconciliation.

What changed

Restored the missing default kataManager values in deployments/gpu-operator/values.yaml.

Validation

I verified the following:

1. Stock v26.3.1 reproduced the issue and the ClusterPolicy reported:

   • Failed to reconcile state-kata-manager: empty image path provided through both ClusterPolicy CR and ENV KATA_MANAGER_IMAGE

2. With this patch applied and the chart rendered locally, the kataManager block was complete again.

3. Upgrading from the patched local chart resulted in:

   • ClusterPolicy.status.state: ready
   • ClusterPolicy is ready as all resources have been successfully reconciled

Notes

For local source-based validation only, I temporarily adjusted Chart.yaml so the raw source chart would not use main-latest, but that change is not part of this PR.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[rajatchopra]

@Ayush-Rathor thanks for this PR.
However, we intend to deprecate KataManager. It is not needed anymore as functionality has moved to cc-manager and sandbox-device-plugin.

the operator hits a reconcile panic in the Kata runtime class path is what we really need to fix here.

[Ayush-Rathor-1]

Got that, just saw your PR for the same