Support speech recognition

[waj]

Google Cloud Speech-to-Text looks like a good fit (https://cloud.google.com/speech-to-text/). AWS also has a speech-to-text service but the non-streaming version of the API is too slow (like 30 seconds to transcribe a single audio) and the streaming API supports only a few languages. Besides that I also found that Google's service is much more flexible and for example it returns digits (0-9) instead of plain text when a number is read.

Add a transcribe command that uses the Google API to convert the audio to text.

Let's expose this command through the speech recognition feature of the Gather command in TwiML: https://www.twilio.com/docs/voice/twiml/gather#input

(new step to use this command with web designer will come later)