Handle new priority param for action forces

[KTrain5169]

The Neuro API was just updated to add a new priority param for actions force (see VedalAI/neuro-sdk@6f473fb)

I would contribute if it wasn't for the Tauri rewrite and also me having... absolutely no clue how this code works 😅

[Govorunb]

That's a cool feature - thanks for the update.

As far as I can make a guess, this has limited impact for gary. I assume it's meant for interrupting speech - waiting for speech to finish is (probably) the thing that takes the vast majority of time for Neuro and we don't do it at all.

Otherwise (just leaving a note for future me) we could interrupt LLM generation - however, I think I remember some part of the Python side making that hard/annoying. So, there's not much I could do to support this at the moment apart from allowing it in the message models. Very much noted though.

[KTrain5169]

Yeah, I think interrupting the LLM generation is something that is the closest simulation to it as possible.