Penalty and front-fixing methods for the numerical solution of American option problems