Efficient Learning of Real Time Two-Counter Automata (Extended Abstract)