Move X-Forwarded-Port variable to the location context

Resolves issue #4038 where the X-Forwarded-Port header would be set to the value of the https listening port if all of the following settings were satisfied:
- The ingress controller was started with a non-default HTTPS port set with the `--https-port` argument
- An ingress is created having:
  - the `nginx.ingress.kubernetes.io/auth-url` annotation set
  - TLS enabled

This commit solves this issue by moving the setting of the `pass_server_port` variable from the server, one level down to the location context.

test/e2e/framework/exec.go

@@ -88,9 +88,9 @@ func (f *Framework) ExecCommand(pod *corev1.Pod, command string) (string, error)
 }
 
 // NewIngressController deploys a new NGINX Ingress controller in a namespace
-func (f *Framework) NewIngressController(namespace string) error {
+func (f *Framework) NewIngressController(namespace string, namespaceOverlay string) error {
 	// Creates an nginx deployment
-	cmd := exec.Command("./wait-for-nginx.sh", namespace)
+	cmd := exec.Command("./wait-for-nginx.sh", namespace, namespaceOverlay)
 	out, err := cmd.CombinedOutput()
 	if err != nil {
 		return fmt.Errorf("Unexpected error waiting for ingress controller deployment: %v.\nLogs:\n%v", err, string(out))