Implement real time audio transcription

I experimented with a demo of real-time speech to text with whisper: https://github.com/davabase/whisper_real_time

In my fork I was able to get it working by pinning the Python dependencies: https://github.com/audreyfeldroy/whisper_real_time/tree/pin-python-dependencies

- [ ] Study and try out the demo code to see how it works
- [ ] Add a command line option to our `hist transcribe` command like `hist transcribe --realtime` to get audio input from a real-time stream instead of a pre-recorded file

Start by studying and running the linked whisper_real_time demo, including the fork's pinned Python dependencies. Then find the existing `hist transcribe` command and determine how its input is handled. Done means a `--realtime` option accepts a real-time audio stream and transcribes it successfully alongside the existing file-based mode.

Written by the indexing model from the issue text.