Learning label == 0 incompatible with iota2

Hi all,

I observed a minor issue when using iota2 for binary classification. While it is common to have labels = [0,1] for this kind of task, this creates some problems for the classification rasters. Indeed, since iota2 considers "0" as no-data, the class 0 should not be used. To avoid any problem, maybe a simple solution could be to raise an error if the class 0 is used.

Regards, Florian