Improve audio device selection

There are a few issues:

Audio devices only scanned on startup.
Audio device has a very prominent place in the UI, which is unnecessary as users will probably select their audio device only once.
Depending on the users latency needs, a different audio API might be better. Currently it uses DirectSound and 'high' latency by default. WDM-KS has much lower latency (~50ms rather than 300ms) but is less robust. Make user selectable?

If you want lower latency now, use:

audiogenerationwidget.py

PREFERRED_HOST_API = WDMKS
PREFERRED_LATENCY = 'low'     # 'low', 'high' or a float

diglet48 / restim