Make up a extensible interface for VT-control. The current code is a mess
and could mess up the system to some extent if used in the wrong way.
When you use the library correctly, this should not happen, but I want it
to be bullet-proof. Please look at the file IDEA for a possible solution
that would not require changing the kernel.

Correctly implement the paged frame-buffer and the corresponding helper 
functions (setpage). This will need another nopage-handler.
Allow for different paging-functions to be called from there. (needed for
16 color modes ...)

> Check whether using functions pointers would give a gain over using the
> "switch" that is worth the probable mess they introduce (maybe find a less
> messy way than that I think of ...).
I tried it and they seem to slow the system down ... please check what could
be improved !

> Check out why acceleration does not work. (Comment it in if you dare - this
> is really a surprising experience ... ) 
> Somebody that knows what (s)he is doing should carefully check the init-
> routine. It must be wrong for accelerated use (it does not work ...),
> but what is wrong ... ???
This will probably be corrected by switching to Steffen's driver when it
works.

