Update to median_abs_deviation from scipy

[d-diaz]

I've noticed that scipy docs indicate that median_absolute_deviation is now deprecated and recommend using median_abs_deviation instead. Doing a calculation of a sample array directly from numpy seems to indicate that the newer version implements the calculation correctly while the older version is different (for some reason I have not investigated).

[sherbold]

Thanks! I will take a look at this in the next days, but this looks like an easy merge.