fix: retry ClusterSummary status update in updateStatusForNonReferencedHelmReleases

[gianlucam76]

In SyncModeContinuousWithDriftDetection, updateStatusForNonReferencedHelmReleases did a single Get plus a single Status().Update with no retry, unlike its two siblings on the same pass (updateStatusForReferencedHelmReleases, updateValueHashOnHelmChartSummary), which both already wrap their Get+Update in retry.RetryOnConflict.

A concurrent status write (e.g. updateValueHashOnHelmChartSummary a few ms earlier) can bump the ClusterSummary's resourceVersion between this function's cached Get and its own Status().Update, so the write loses with:

Operation cannot be fulfilled on clustersummaries.config.projectsveltos.io "...":
the object has been modified; please apply your changes to the latest version and try again

handleCharts returns that error immediately, even though the Helm deploy already succeeded. Consequences while this keeps happening:

• the feature hash never advances, so every reconcile re-runs a full no-op deploy pass
• the Helm feature reports Failed with consecutiveFailures climbing, while the release itself is healthy and deployed
• drift detection registration (postProcessDeployedHelmCharts) is never reached, so ResourceSummary.spec.chartResources stays empty and out-of-band drift goes undetected

This mostly hits profiles under frequent reconcile pressure (e.g. an HPA-autoscaled Deployment keeps requesting reconciliation), since each failed pass's own status write (consecutiveFailures) supplies the next conflicting write, making the loop self-sustaining once triggered.

Fix: wrap the Get and Status().Update in updateStatusForNonReferencedHelmReleases in retry.RetryOnConflict, matching its two siblings exactly.

Fixes #1933