core: allocate control structures using an own kmem_cache

Testing /dev/urandom versus zzero-0-1-data (the random device)
shows this increases performance by 0.1 usecs per block. We
are now just over 0.7 usecs per block on my development PC.

This also introduces zio-core.c with generic stuff. We agree that zio-sys
will become zio-sysfs over time.

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Conflicts:

	Makefile