Handle no available nodes to cluster with

This came up when testing whether all nodes would wait for node 0
could not start. To test this, I was killing node 0 every few seconds,
so it would attempt to start and then get killed. This led to other
nodes sometimes discovering it and attempting to sync, but then crashing
with function_clause as it went away.

deps/rabbit/src/rabbit_peer_discovery.erl

@@ -835,6 +835,12 @@ can_use_discovered_nodes(_DiscoveredNodes, []) ->
 %%
 %% @private
 
+select_node_to_join([]) ->
+    ?LOG_INFO(
+       "Peer discovery: no nodes available for auto-clustering; waiting before retrying...",
+       [],
+       #{domain => ?RMQLOG_DOMAIN_PEER_DISC}),
+    false;
 select_node_to_join([{Node, _Members, _StartTime, _IsReady} | _])
   when Node =:= node() ->
     ?LOG_INFO(