Move the peer manager to be a behaviour (#2773)

This simply moves some functions that were "swarm notifications" to a network behaviour implementation.

Notes
------
- We could disconnect from the peer manager but we would lose the rpc shutdown message
- We still notify from the swarm since this is the most reliable way to get some events. Ugly but best for now
- Events need to be pushed with "add event" to wake the waker

Co-authored-by: Divma <26765164+divagant-martian@users.noreply.github.com>

beacon_node/http_api/tests/common.rs

@@ -6,6 +6,7 @@ use eth2::{BeaconNodeHttpClient, Timeouts};
 use http_api::{Config, Context};
 use lighthouse_network::{
     discv5::enr::{CombinedKey, EnrBuilder},
+    libp2p::{core::connection::ConnectionId, swarm::NetworkBehaviour},
     rpc::methods::{MetaData, MetaDataV2},
     types::{EnrAttestationBitfield, EnrSyncCommitteeBitfield, SyncState},
     ConnectedPoint, Enr, NetworkGlobals, PeerId, PeerManager,
@@ -118,8 +119,8 @@ pub async fn create_api_server<T: BeaconChainTypes>(
         local_addr: EXTERNAL_ADDR.parse().unwrap(),
         send_back_addr: EXTERNAL_ADDR.parse().unwrap(),
     };
-    let num_established = std::num::NonZeroU32::new(1).unwrap();
-    pm.inject_connection_established(peer_id, connected_point, num_established, None);
+    let con_id = ConnectionId::new(1);
+    pm.inject_connection_established(&peer_id, &con_id, &connected_point, None);
     *network_globals.sync_state.write() = SyncState::Synced;
 
     let eth1_service = eth1::Service::new(eth1::Config::default(), log.clone(), chain.spec.clone());