Namespace changes in 1.1.0 breaks upgrades from 1.1.0 if not using default namespaces

[therealdwright]

/kind bug

What happened?
When upgrading to 1.1.0 from 1.0.0 all ClusterRoleBinding's and RoleBinding's were shifted in to the kube-system namespace while our deployment which references the upstream overlays is deployed to a namespace other than kube-system. This lead to relevant RBAC permissions being inaccessible to the ServiceAccount's that rely on them.

What you expected to happen?
Kustomize should be able to decide which namespace it deploys the aws-ebs-csi-driver to ensuring resources are not split between multiple namespaces.

How to reproduce it (as minimally and precisely as possible)?

1. Deploy aws-ebs-csi-driver to a non-default namespace using the base below.

apiVersion: kustomize.config.k8s.io/v1beta1
kind: Kustomization
resources:
  - github.com/kubernetes-sigs/aws-ebs-csi-driver/deploy/kubernetes/overlays/stable?ref=8b1e568654937ac42ff750fbee79b207d1e63016
namespace: foo
 

2. Upgrade to 1.1.0 by using the base below.

apiVersion: kustomize.config.k8s.io/v1beta1
kind: Kustomization
resources:
  - github.com/kubernetes-sigs/aws-ebs-csi-driver/deploy/kubernetes/overlays/stable?ref=f0011c9fc23afbcea3bd8af97627da8ad3dc1e84
namespace: foo
 

Anything else we need to know?:

Environment

• Kubernetes version (use kubectl version):
  Client Version: v1.21.0
  Server Version: v1.20.4-eks-6b7464
• Driver version:
  1.0.0 upgrading to 1.1.0

[therealdwright]

Introduced in #878