Fix panic when checking service status (elastic#1078)

There seems to be a race condition since a container starts running,
till its state reports any health information. During this time the
value obtained from the API is nil. Don't try to get the status from
the healtcheck if that's the case.

Author: jsoriano

internal/stack/compose.go

@@ -221,7 +221,11 @@ func newServiceStatus(description *docker.ContainerDescription) (*ServiceStatus,
 		Version: getVersionFromDockerImage(description.Config.Image),
 	}
 	if description.State.Status == "running" {
-		service.Status = fmt.Sprintf("%v (%v)", service.Status, description.State.Health.Status)
+		healthStatus := "unknown health"
+		if health := description.State.Health; health != nil {
+			healthStatus = health.Status
+		}
+		service.Status = fmt.Sprintf("%v (%v)", service.Status, healthStatus)
 	}
 	if description.State.Status == "exited" {
 		service.Status = fmt.Sprintf("%v (%v)", service.Status, description.State.ExitCode)