On the Complexity of Rational Verification