Always shutdown etcd when stopping envtest

This patch fixes an issue that occurs when shutting down an
envtest.Environment. If the API server fails to shutdown, then the etcd
process is orphaned. This can result in dozens of running etcd processes
when debugging failed tests.

Please see this Slack discussion for more information:
https://kubernetes.slack.com/archives/CAR30FCJZ/p1590779385006000.

Author: akutz

pkg/internal/testing/integration/control_plane.go

@@ -5,6 +5,7 @@ import (
 	"net/url"
 
 	"k8s.io/apimachinery/pkg/runtime/serializer"
+	utilerrors "k8s.io/apimachinery/pkg/util/errors"
 	"k8s.io/client-go/kubernetes/scheme"
 	"k8s.io/client-go/rest"
 
@@ -42,17 +43,20 @@ func (f *ControlPlane) Start() error {
 
 // Stop will stop your control plane processes, and clean up their data.
 func (f *ControlPlane) Stop() error {
+	var errList []error
+
 	if f.APIServer != nil {
 		if err := f.APIServer.Stop(); err != nil {
-			return err
+			errList = append(errList, err)
 		}
 	}
 	if f.Etcd != nil {
 		if err := f.Etcd.Stop(); err != nil {
-			return err
+			errList = append(errList, err)
 		}
 	}
-	return nil
+
+	return utilerrors.NewAggregate(errList)
 }
 
 // APIURL returns the URL you should connect to to talk to your API.