The async_recognize tests need to use an instance of `SpeechContext`. See: https://github.com/GoogleCloudPlatform/google-cloud-python/pull/2680#discussion_r86839523