Knowledge of Catalan, public/private sector choice and earnings: Evidence from a double sample selection model