uterm: video: modularize video backends

uterm: video: modularize video backends

We want to allow dynamic backend loading during runtime so we need to
avoid static definitions. This converts the backend IDs into module
structures which are then linked into the binary.

To make this fully modular, we need to load the backends dynamicly via
dlsym() and move all the static declarations from uterm_video.h into the
backends. But for now at least the backends can be easily exchanged and
configured during compile-time.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>