Do parallel transcription with Deep Speech

[ianb]

I believe there's a request flag to do Deep Speech transcription. That is, with just a small tweak we could use Deep Speech.

While I'm not sure we want to switch entirely (though we could make that an option), we could send the request twice, get both responses, and record both transcriptions in Telemetry (and a flag just for equality, which wouldn't require opt-in). This probably doesn't require any server changes.

[awallin]

Would be great to collect transcriptions for both during internal Beta and do a comparison to evaluate quality.

[ianb]

I'm asking @andrenatal about what we have to change about the request to change the transcription backend.

[andrenatal]

@ianb just replace the service endpoint by: https://dev.speaktome.nonprod.cloudops.mozgcp.net

This is a backend with deep speech

[ianb]

mozilla-services/mozilla-pipeline-schemas#444 contains new Telemetry properties related to this.