Can spoofing countermeasure and speaker verification systems be jointly optimised?