On the computational content of the Loeb measure