Knowledge of Catalan, Public/Private Sector Choice and Earnings: Evidence from a Double Sample Selection Model