feat: new proving broker implementation

[alexghr]

This PR adds ProvingBroker which implements a new interface for distributing proving jobs to workers as specified in #8495

[spalladino]

Looking good! Most of my comments are around having a different prior mental model for the broker and how it interacts with the backend. I had assumed the broker would keep all metadata needed in memory, and use a backend to persist it and reload in the event of a crash, and another backend for storing and transmitting payloads. Does that make sense?

[spalladino]

I think we independently introduced the same new types :-P

[alexghr]

use a backend to persist it and reload in the event of a crash, and another backend for storing and transmitting payloads. Does that make sense?

This makes sense! Initially I only thought about the inputs needing to be stored separately, but you're right it makes sense to store inputs and outputs to/from the proving process.

I the next PR I'll change the ProvingJob and ProvingJobResult types to take URIs to inputs and proofs/public inputs instead of the actual data structures.