Use RpcSend on RPC::self_limiter::ready_requests

[jxs]

Issue Addressed

We don't need to store BehaviourAction for ready_requests and therefore avoid having an unreachable! on #6625.
Therefore this PR should be merged before it

[ackintosh]

Thanks for this @jxs! This is helpful for simplifying #6625, but do we still need unreachable! on #6625 like this?:

                if let RPCSend::Request(request_id, req) = event {
                    //
                } else {
                    unreachable!()
                }

[jxs]

Thanks for this @jxs! This is helpful for simplifying #6625, but do we still need unreachable! on #6625 like this?:

                if let RPCSend::Request(request_id, req) = event {
                    //
                } else {
                    unreachable!()
                }

we can just return false on that situation right? (same as before true)

[ackintosh]

It will look like this if we return false instead of unreachable! on #6625:

    pub fn peer_disconnected(&mut self, peer_id: PeerId) -> Vec<(Id, Protocol)> {
        // ...

        self.ready_requests.retain(|(req_peer_id, event)| {
            if let RPCSend::Request(request_id, req) = event {
                if req_peer_id == &peer_id {
                    failed_requests.push((*request_id, req.protocol()));
                    // Remove the entry
                    false
                } else {
                    // Keep the entry
                    true
                }
            } else {
                // unreachable!()

                // Remove the unexpected RPCSend variant 
                false
            }
        });

We can do that, but I think unreachable! is more appropriate in this situation to clearly indicate that the else clause is unreachable.