Hi team,

Opening this bug as the operator is deleting some annotations that I set up on the `knative-serving` namespace.

Use-Case: Add the annotation `openshift.io/node-selector: "tier=infra"` to the `knative-serving` namespace in order to deploy all the control-plane pods in the nodes tagged as `tier=infra` [1]

Versions:

• Openshift: 4.6
• Serverless-operator: 1.14
• knative-serving: 0.20.0

How to reproduce: These are the manifest that I'm using to deploy:

Let's assume the Serverless-operator is already installed

1. Create the `knative-serving` namespace

—
 apiVersion: v1
 kind: Namespace
 metadata: 
 name: knative-serving
 annotations:
 "openshift.io/node-selector": "tier=infra"
 labels:
 "openshift.io/cluster-monitoring": "true"

2. Deploy the `knative-serving` CRD:

apiVersion: operator.knative.dev/v1alpha1
kind: KnativeServing
metadata:
 name: knative-serving
 namespace: knative-serving
spec:
 version: stable
 controller-custom-certs:
   name: config-service-ca
   type: ConfigMap
 high-availability:
   replicas: 1

What I expect: All the related pods are created on the nodes tagged as `tier=infra`

What I get: The annotation `openshift.io/node-selector: "tier=infra"` is deleted from the namespace when I apply the CRD manifest, and the pods are created on all the available nodes.

[1]: https://docs.openshift.com/container-platform/4.6/nodes/scheduling/nodes-scheduler-node-selectors.html