Data model for transcripts

[Frando]

Currently the transcript is saved as JSON with the following structure:

{
   "parts": [
         { "word": "foo", "conf": 0.382, "start": 1.323, "end": 1.55322 }
    ]
}

Missing things:

• If we continue to store the full transcript results, we need an optional field per part "suffix" to store the punctuation that comes afterwards
• For elasticsearch indexing we already convert this format into a text-only form for the elasticsearch payload delemiter. The format is `"foo|start,end,conf bar|start,end,conf". This is much smaller. Possibly switch to this format for storage as well.
• We should also define an outer data model for transcript to store meta information. Fields: engine, model, modelVersion, (and processingTime, averageConfidence, createdDate - can be autogenerated).