-------------------------------------------
variable: $fillpix -- 3d fill pixel size
-------------------------------------------
default: 1.0

If this is set to the current pixel size (1.0 for
a 256^3 dataset or 0.5 for a 512^3 dataset), a 3D
fill (FILL3d button, to right) will be performed
with native-sized pixels resulting in a volume
similar (but not pixelwise identical) to that
produced by fill.

If fillpix is set larger than native, this will
result in a smaller number of total voxels in the
fill.  The 3D dataset will have the same bounding
box and the same pixel width, but the brain will
be scaled down.

The effect is equivalent to scaling down the
surface by 1.0/(fillpix/nativepix) (e.g., with
the function really_scale_brain) and filling with
native-sized voxels.

Setting fillpix larger than native this way can
be used to downsample a surface.  The resulting
fill.$hemi.mgz can be directly used as an input
to make a surface, which will have fewer vertices
than the original surface but be the same shape.
It will be smaller in linear dimensions by the
fraction: fillpix/nativepix.

N.B.: setting fillpix smaller than native voxels
(in order to upsample) isn't working (yet).

