Per Pippin, the only color model that can
have double precision is RGB/A.
Therefore, we need to switch all others to
use float instead. This patch converts the
HSV and HSVA double babl formats.