Return indices of the minimum values along the given axis.

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