Add an option to generate noise during speech-dispatcher execution

Some headphones or sound system go idle when no sound is produced. This means that the beginning of all sentences are eaten by the wake-up of the sound system.

It would be useful to add an option to speech-dispatcher to permanently generate noise, so as to keep such systems up and running. The volume of the noise should be configurable, to make it loud enough to avoid the system going idle, but silent enough to avoid user headache.

The implementation could be a mere thread started during speech-dispatcher startup, that just generates the noise and pushes it to an instance of audio output.

No files or tests are named. Start by tracing speech-dispatcher startup and the audio output instance, then determine where a background noise generator and configurable volume belong; done means the option keeps supported sound systems awake without disrupting speech.

Written by the indexing model from the issue text.