exit 5 instead of raising error if no etcd endpoint responds (#886)

Exit 5 (5th letter, E, for etcd) instead of raising error when no etcd endpoint responds.

fixes #872

spec/etcd_spec.cr

@@ -27,7 +27,7 @@ describe LavinMQ::Etcd do
         etcd.watch("foo") do |val|
           w.send val
         end
-      rescue LavinMQ::Etcd::Error
+      rescue SpecExit
         # expect this when etcd nodes are terminated
       end
       w.receive # sync
@@ -51,15 +51,15 @@ describe LavinMQ::Etcd do
         etcd.elect_listen(key) do |value|
           leader.send value
         end
-      rescue LavinMQ::Etcd::Error
+      rescue SpecExit
         # expect this when etcd nodes are terminated
       end
       lease = etcd.elect(key, "bar", 1)
       leader.receive.should eq "bar"
       spawn(name: "elect other leader spec") do
         begin
           etcd.elect(key, "bar2", 1)
-        rescue LavinMQ::Etcd::Error
+        rescue SpecExit
           # expect this when etcd nodes are terminated
         end
       end

src/lavinmq/etcd.cr

@@ -261,7 +261,8 @@ module LavinMQ
         Log.debug { "Could not connect to #{address}: #{ex}" }
         next
       end
-      raise Error.new("No endpoint responded")
+      Log.fatal { "No etcd endpoint responded" }
+      exit 5 # 5th character in the alphabet is E(etcd)
     end
 
     private def update_endpoints(tcp, address)