Avoid using errors to communicate the state of other shards to the leader during query-status API.

[cberkhoff]

Before this change, when the leader shard received a query-status API, it would call a match-status API on its shards, which would return an error if the state was different. The error contained the status of each shard, that the leader would use to compute the overall helper state. This wasn't the intended behavior, and was only done as a stop-gap solution, because the Transport::send (and consequently Transport::broadcast) returned unit "()", prohibiting from receiving the state from other shards. This PR changes this:

• Transport::send now returns a configurable associated type for all requests (Transport::SendReponse)
• MPC Query Status HTTP server handle was re-used as a Shard API. Match-status API is gone.
• HTTP client query-status API can return either a QueryStatus object for compatibility or a BodyStream.
• Processor now simply broadcast query-status API and compiles the results. Errors are handles as before. Previously used helper functions to cast the errors are gone. Tests were updated.
• Added some calls to query-status in HTTP tests, to make sure the parsing is fine and that the changes work end-to-end.

I also tried a bunch of things so that Transport::send would return a specific type, bound to the route, but it all seemed very complex. The goal was that QueryStatus route would return a QueryStatusResponse, the rest could just return unit. That task can be tackled separately in the future.