fix registry addon for containerd runtime.

[medyagh]

Our Registry addon is not working on Containerd

https://storage.googleapis.com/minikube-builds/logs/10758/9ba6c7c/Docker_Linux_containerd.html#fail_TestAddons%2fparallel%2fRegistry

=== RUN   TestAddons/parallel/Registry
=== PAUSE TestAddons/parallel/Registry
=== CONT  TestAddons/parallel/Registry
=== CONT  TestAddons/parallel/Registry
addons_test.go:211: registry stabilized in 24.400968ms
=== CONT  TestAddons/parallel/Registry
addons_test.go:213: (dbg) TestAddons/parallel/Registry: waiting 6m0s for pods matching "actual-registry=true" in namespace "kube-system" ...
=== CONT  TestAddons/parallel/Registry
helpers_test.go:335: "registry-x65s9" [d8e5005c-9d91-4bb8-9192-f1103d4c6fba] Running
=== CONT  TestAddons/parallel/Registry
addons_test.go:213: (dbg) TestAddons/parallel/Registry: actual-registry=true healthy within 5.008994314s
=== CONT  TestAddons/parallel/Registry
addons_test.go:216: (dbg) TestAddons/parallel/Registry: waiting 10m0s for pods matching "registry-proxy=true" in namespace "kube-system" ...
helpers_test.go:335: "registry-proxy-vsjkn" [d8202aa1-45d7-4bfe-aa19-54fc9315297d] Running
=== CONT  TestAddons/parallel/Registry
addons_test.go:216: (dbg) TestAddons/parallel/Registry: registry-proxy=true healthy within 5.010342896s
addons_test.go:221: (dbg) Run:  kubectl --context addons-20210310185414-7328 delete po -l run=registry-test --now
addons_test.go:226: (dbg) Run:  kubectl --context addons-20210310185414-7328 run --rm registry-test --restart=Never --image=busybox -it -- sh -c "wget --spider -S http://registry.kube-system.svc.cluster.local"
=== CONT  TestAddons/parallel/Registry
addons_test.go:226: (dbg) Non-zero exit: kubectl --context addons-20210310185414-7328 run --rm registry-test --restart=Never --image=busybox -it -- sh -c "wget --spider -S http://registry.kube-system.svc.cluster.local": exit status 1 (1m0.141024647s)
-- stdout --
	pod "registry-test" deleted
-- /stdout --
** stderr ** 
	error: timed out waiting for the condition
** /stderr **
addons_test.go:228: failed to hit registry.kube-system.svc.cluster.local. args "kubectl --context addons-20210310185414-7328 run --rm registry-test --restart=Never --image=busybox -it -- sh -c \"wget --spider -S http://registry.kube-system.svc.cluster.local\"" failed: exit status 1
addons_test.go:232: expected curl response be "HTTP/1.1 200", but got *pod "registry-test" deleted
*
addons_test.go:240: (dbg) Run:  out/minikube-linux-amd64 -p addons-20210310185414-7328 ip
2021/03/10 18:58:41 [DEBUG] GET http://192.168.49.161:5000
addons_test.go:269: (dbg) Run:  out/minikube-linux-amd64 -p addons-20210310185414-7328 addons disable registry --alsologtostderr -v=1

[afbjorklund]

It was also broken for CRI-O, and had been broken for Docker too - if it had used CRI.