feature request for Binner

Hi,

When running below ImageJ macro...

```
newImage("Untitled", "8-bit black", 256, 254, 11);
run("Bin...", "x=3 y=3 z=3 bin=Average");
```

... the resulting image has 3 z-planes, indicating that the Binner discards the information in slices 10 and 11 of the original data. One other choice of implementing the Binner would have been that the resulting image has 4 z-planes, where its 4th z-plane is computed from the information in planes 10 and 11 of the original data.

I feel both options are valid, for me personally at the moment the latter option would be better, in order to preserve all content of my data.

Would it be possible to add an option such that the user can choose between the two behaviors?

Start by locating the Binner implementation and its handling of the ImageJ macro parameters for x, y, z, and bin=Average. Run the supplied macro to reproduce the three-plane result, then determine how the alternate handling of leftover slices should be exposed and verified. Done means the user can choose whether leftover slices are discarded or combined into an additional output plane.

Written by the indexing model from the issue text.