drm/mgag200: Fix LUT programming for 16bpp

Since there are only 32 (64) distinct color values for each color
in 16bpp Matrox hardware expects those in a 'dense' manner, ie in
the first 32 (64) entries of the respective color.

Signed-off-by: Egbert Eich <eich@suse.de>
Signed-off-by: Dave Airlie <airlied@redhat.com>