fix: nvca-operator chart clusterSource defaults to self-managed instead of ngc-managed in 3.1.0

[apartha-nv]

Problem

Two bugs affect nvca-operator 3.1.0 upgrades:

Bug 1 - Wrong clusterSource default

The 3.1.0 chart bakes ngcConfig.clusterSource: self-managed as the default. The correct
default for the public chart is ngc-managed. NGC UI-generated install commands do not pass
--set ngcConfig.clusterSource=..., so any cluster installed with the 3.1.0 chart without
an explicit override never registers with the NVCA agent.

Bug 2 - Nil pointer on upgrade from 3.0.x

deploy/helm/nvca-operator/nvca-operator/templates/self-managed-nvcfbackend-cm.yaml evaluates
.Values.selfManaged.imageCredHelper.imageRepository and .Values.selfManaged.sharedStorage.imageRepository
unconditionally in the ConfigMap metadata section before the clusterSource guard on line 37.
These sub-keys were added in 3.1.0. When upgrading from 3.0.x with --reuse-values, the stored
values do not contain these sub-keys, making their parent maps nil. Accessing .imageRepository
on a nil interface panics with: nil pointer evaluating interface {}.imageRepository.

This panic affects all cluster types (including ngc-managed) because the metadata block runs
regardless of clusterSource.

Fix

• Restore ngcConfig.clusterSource: ngc-managed as the chart default in both the vendoring
  script and the published values.yaml.
• Add ngcConfig.clusterSource: self-managed to the self-managed stack helmfile so that
  self-hosted installs continue to work correctly.
• Use | default dict before accessing .imageRepository on selfManaged.imageCredHelper
  and selfManaged.sharedStorage to guard against nil parents on upgrade.