Don't recursively call deregister_worker() on the current worker

Previously we were not filtering out the current worker when calling
`deregister_worker()` on `workers()`.

Author: JamesWrigley

src/cluster.jl

@@ -1185,7 +1185,7 @@ function deregister_worker(pg, pid)
             # Notify the cluster manager of this workers death
             manage(w.manager, w.id, w.config, :deregister)
             if PGRP.topology !== :all_to_all || isclusterlazy()
-                for rpid in workers()
+                for rpid in filter(!=(myid()), workers())
                     try
                         remote_do(deregister_worker, rpid, pid)
                     catch