Part of fixing bug 5073407 was to upgrade the interface used by our
ScaledBlit primitives to provide sub-pixel positioning information.

Those changes introduced some new paradigms in how we pass around
image coordinates in our image pipelines which conflicted with a
large amount of the OpenGL pipeline code. Rather than complicate
the code changes to fix 5073407 with lots of tiny code changes to
the OpenGL code, only the upper level interfaces were changed as
necessary to integrate with the new public calling conventions and
the internal code paths were left in the old style.

Additionally, the new parameter lists involve double-precision
floating point parameters, but the library entry points into the
OpenGL library for the double precision numbers were never imported.
Currently the code simply casts down from double to float to use
the single-precision entry points, but the double-precision entry
points should be used for maximum benefit.