Fix rpl_events test failure in the runtime tree.

mysql-test/r/rpl_events.result:
  Now ON COMPLETION NOT PRESERVE events are also dropped on the
  slave, since DROP EVENT command that is invoked for all such commands
  gets invoked on the slave.
sql/event_data_objects.cc:
  Fix the failing rpl_events test after the patch for Bug#27733.
  At the time Events::drop_event got invoked inside
  Event_job_data::execute() thd->query pointed to CREATE PROCEDURE
  statement. This statement was written to the binary log
  from inside Events::drop_event (under assumption that this is a
  DROP EVENT statement that needs to be replicated), and caused
  creation of this procedure on the slave (and a subsequent failure
  when a procedure with the same name already exist).
  
  The patch ensures that thd->query points at the right query text
  for DROP EVENT executed when dropping ON COMPLETION NOT PRESERVE 
  events.
sql/event_data_objects.h:
  Update a  declaration.
sql/events.cc:
  Change if () to an assert: thd->query now always points at a valid
  query.

sql/event_data_objects.h

@@ -184,6 +184,8 @@ public:
 private:
   bool
   construct_sp_sql(THD *thd, String *sp_sql);
+  bool
+  construct_drop_event_sql(THD *thd, String *sp_sql);
 
   Event_job_data(const Event_job_data &);       /* Prevent use of these */
   void operator=(Event_job_data &);