Draft: Scalable architecture design

A proposal for a new scalable architecture. Feedback is requested.

If not apparent, this is my clumsy attempt at porting the idea of hybrid threading onto the service level. Please see docs/internal/architecture.md for the meat of the actual design. The implementation details are intentionally omitted for now.

The high level overview also doesn't really cover any of the interesting bits around failure handling and the scheduling part yet. I will make a draft for these after my absence and likely delegate part of the load balancing to my BA, if possible.