feat(rust): improve ockam_node

[SanjoDeundiak]

1. Using Oneshot channels where possible - faster, better semantics, avoiding async when sending a signal
2. Make start_worker and stop_worker sync, which allowed to make a lot of things sync, like start a secure channel listener, outlet, create TcpTransport, etc.
3. Removing code that was leading to implicit Clones (like &str -> Address conversion). Cloning causes a lot of small allocation and memory fragmentation => therefore should be avoided => therefore should explicit for starters.
4. HashMaps and HashSet instead of BTreeMap and BTreeSet for better performance
5. More granular locks (lock individual fields instead of the whole struct) inside Router.
6. Renaming to improve vocabulary consistency:
   • ctx.stop_worker(), ctx.stop_processor() -> ctx.stop_address()
   • ctx.stop() -> ctx.shutdown_node()
   • node.stop() -> node.shutdown()
   • terminate was used together with stop for the same things -> removed
7. Using Weak<Router> for clearer ownership over Router, which helps to drop it earlier and in more predictable way. E.g., detached Context will no longer keep Router from running => we can shutdown the environment successfully and then detached Context instances that are left can be dropped safely considering they no longer has access to the Router.
8. Use tokio::Handle instead of tokio::Runtime where possible, to achieve much clearer ownership over tokio's Runtime. Owning the Runtime instance keeps runtime alive + exclusive right to terminate it, while tokio::Handle only gives access to the Runtime, but doesn't keep it alive itself.
9. Got read of AsyncTryClone. Cloning Context type became sync, so we no longer need AsyncTryClone. Still needed failable clone, so replaced with TryClone
10. Made Worker and Processor treatment by the Router/Node almost identical, which will ease future merge of this abstractions.
11. Replaced clusters with Priority shutdown. Each worker can specify its shutdown priority (or get a default one). During node shutdown workers will be stopped in batches starting with the highest priority to the lowest. Why it's useful: workers rely on other workers, e.g. secure channel worker needs tcp connection worker to deliver messages. Stopping secure channel worker first, can possible allow it to still use tcp connection worker to deliver some final message(s) to the other side (e.g., "close secure channel message")

Possible future improvements:

1. More tests
2. Merge Worker and Processor into one trait
3. Try lockless structures (HashMaps and HashSets at least)
4. Shutdown tokio Environment in cases where something hangs and can't shutdown gracefully

[SanjoDeundiak]

@etorreborre is it ok to delete it? It looks unused

[davide-baldo]

Looks good!
Everything can be addressed in a follow-up

[davide-baldo]

Was this a bug?

[SanjoDeundiak]

Could you please elaborate?

[davide-baldo]

You removed if result.is_some() {, and result comes from the previous select!{} condition result = self.recv_message() => {, was this preventing a relay shutdown?