Return indices of the maximum values along the given axis.

This is similar to argmax(), but adapted to ensure that the full precision given by the two doubles jd1 and jd2 is used. See argmax() for detailed documentation.