Abort add_node when stale replication origin cleanup fails

Stale-origin drop exists to prevent create_sub from reusing a leftover
LSN and silently skipping data.  Swallowing the error with RAISE WARNING
defeated that safety net.

Author: mason-sharp

samples/Z0DAN/zodan.sql

@@ -1507,7 +1507,7 @@ BEGIN
                     slot_name;
             EXCEPTION
                 WHEN OTHERS THEN
-                    RAISE WARNING '    Could not drop stale origin % on new node: %',
+                    RAISE EXCEPTION 'Could not drop stale origin % on new node: %',
                         slot_name, SQLERRM;
             END;
             CALL spock.create_sub(