Do not truncate the list of peers for propagation

The vector obtained from Topology::view is sorted by node ID,
so truncations make the network biased toward nodes with low IDs.
Let the client connections rotate "naturally", allowing some churn.
A limit on the view size should be implemented in Poldercast:
primetype/poldercast#6

jormungandr/src/network/mod.rs

@@ -351,7 +351,6 @@ fn handle_propagation_msg(
     // the item.
     send_to_peers.then(move |res| {
         if let Err(mut unreached_nodes) = res {
-            unreached_nodes.truncate(state.config.max_client_connections);
             debug!(
                 state.logger(),
                 "will try to connect to {} of the peers not immediately reachable for propagation",
@@ -418,7 +417,6 @@ fn send_gossip(state: GlobalStateR, channels: Channels) -> impl Future<Item = ()
         .view(poldercast::Selection::Any)
         .and_then(move |view| {
             let mut peers = view.peers;
-            peers.truncate(state.config.max_client_connections);
             debug!(logger, "sending gossip to {} peers", peers.len());
             stream::iter_ok(peers).for_each(move |node| {
                 let peer_id = node.id();