(bug) handle helm chart errors in pullmode

[gianlucam76]

Previously, when Sveltos processed a Helm chart in pull mode, it would generate ConfigurationBundle instances as it prepared the resources. If an error was encountered halfway through this process, Sveltos would still commit the resources it had successfully prepared up to that point.

This led to a critical synchronization failure:

1. Partial State: The ConfigurationBundle contained only a subset of the intended resources.
2. Applier Misinterpretation: The applier would treat the missing resources as "deleted" or "out of scope," causing it to prune existing resources on the managed cluster or deploy an incomplete, broken stack.

This PR introduces atomic preparation logic. If an error occurs at any point during the preparation of resources for a Helm chart:

1. Discard: All partially prepared resources are discarded.
2. No Commit: Nothing is committed to the ConfigurationBundle, ensuring the applier does not act on incomplete data.
3. Report: The error is captured and reported specifically within the ClusterSummary status for visibility.

Fixes #1724

[gianlucam76]

ClusterSummary Status

    featureSummaries:
    - consecutiveFailures: 35
      failureMessage: |
        values don't meet the specifications of the schema(s) in the following chart(s):
        cert-manager:
        - at '/crds': additional properties 'replicas' not allowed
      featureID: Helm
      hash: B2G6wtbQhKUtF0v+KHC5r7qxVhltPre/I36fvFGtwMk=
      lastAppliedTime: "2026-04-18T20:28:36Z"
      status: Failed
    helmReleaseSummaries:
    - failureMessage: |
        values don't meet the specifications of the schema(s) in the following chart(s):
        cert-manager:
        - at '/crds': additional properties 'replicas' not allowed
      releaseName: cert-manager
      releaseNamespace: cert-manager
      status: Managing
      valuesHash: uHgnUljrEBsWynaXJjcLt0c5djSe1eNQ7xiaoopttaE=
    nextReconcileTime: "2026-04-18T20:29:36Z"

all resources are still on the managed cluster