[TASK-659] [TASK-556] Add NLP support for background-audio questions

[jnm]

Checklist

1. If you've added code that should be tested, add tests
2. If you've changed APIs, update (or create!) the documentation
3. Ensure the tests pass
4. Make sure that your code lints and that you've followed our coding style
5. Write a title and, if necessary, a description of your work suitable for publishing in our release notes
6. Mention any related issues in this repository (as #ISSUE) and in other repositories (as kobotoolbox/other#ISSUE)
7. Open an issue in the docs if there are UI/UX changes

Description

Adds back-end support for transcripts, translations, and qualitative analysis of background-audio questions, along with a minimal, proof-of-concept commit to enable this in the front end's single processing view.

Notes

This is a draft PR because background-audio having a dash in it necessitates dealing with qpath using dash as a delimiter; see https://www.notion.so/kobotoolbox/Pick-different-delimiter-for-qpath-probably-with-URL-encoding-e1271ab8ec4442bcb7c5af231f751127.

Internal discussion is at https://chat.kobotoolbox.org/#narrow/stream/4-Kobo-Dev/topic/NLP.20improvements/near/380367.

Related issues

Part of #4407

[notion-workspace]

Cannot transcribe background audio collected by KoboCollect [Backend]

[notion-workspace]

Cannot transcribe background audio collected by KoboCollect [Frontend]