controller: resolve unavailable in conjuction with surge for deployments

Due to rounding down for maxUnavailable, we may end up with deployments
that have zero surge and unavailable pods something that 1) is not allowed
as per validation, 2) blocks deployments. If we end up in such a situation
set maxUnavailable to 1 on the theory that surge might not work due to
quota.

pkg/controller/deployment/deployment_controller_test.go

@@ -230,13 +230,14 @@ func TestDeploymentController_reconcileOldReplicaSets(t *testing.T) {
 		expectedOldReplicas int
 	}{
 		{
-			deploymentReplicas: 10,
-			maxUnavailable:     intstr.FromInt(0),
-			oldReplicas:        10,
-			newReplicas:        0,
-			readyPodsFromOldRS: 10,
-			readyPodsFromNewRS: 0,
-			scaleExpected:      false,
+			deploymentReplicas:  10,
+			maxUnavailable:      intstr.FromInt(0),
+			oldReplicas:         10,
+			newReplicas:         0,
+			readyPodsFromOldRS:  10,
+			readyPodsFromNewRS:  0,
+			scaleExpected:       true,
+			expectedOldReplicas: 9,
 		},
 		{
 			deploymentReplicas:  10,
@@ -492,11 +493,12 @@ func TestDeploymentController_scaleDownOldReplicaSetsForRollingUpdate(t *testing
 		expectedOldReplicas int
 	}{
 		{
-			deploymentReplicas: 10,
-			maxUnavailable:     intstr.FromInt(0),
-			readyPods:          10,
-			oldReplicas:        10,
-			scaleExpected:      false,
+			deploymentReplicas:  10,
+			maxUnavailable:      intstr.FromInt(0),
+			readyPods:           10,
+			oldReplicas:         10,
+			scaleExpected:       true,
+			expectedOldReplicas: 9,
 		},
 		{
 			deploymentReplicas:  10,