Extend One Word support to also send 'finished speaking' events

[ciaranj]

To better improve the 'stateful' understanding of /what/ willow is doing, and for how long. It would be useful if Willow could tell WAS 'I've finished responding'.

WAS currently knows, wake_start, wake_end, notify_done, when a command starts and when a command ends. This allows command endpoints to track 'what a given willow device is doing'.

There is a gap though; While we can infer 'response starting' (as that correlates with 'command end', we can't currently determine 'response completed', the point at which Willow 'stops speaking' / has implemented whatever response activity was needed (there exist non-speaking results)

Currently the wake_start stuff is intrinsically linked to the one wake (WOW) functionality. We could either decide that WAS 'knowing' what willow is up to, is always important, and remove the dependency on WOW or add this new event, if and only if, WOW is enabled. I'm not fussed, although the less 'advanced' settings we need to switch on, the more 'out of the box' we can be.

[stintel]

Should be relatively easy to hook this up in the ESP Audio callback:

https://github.com/HeyWillow/willow/blob/main/main/audio.c#L88-L94

And we could unconditionally send wake_start/wake_end, end instead if guard handling the wake_result response if One Wake is disabled.