generic: per-device coherent dma allocator

Currently x86_32, sh and cris-v32 provide per-device coherent dma
memory allocator.

However their implementation is nearly identical. Refactor out
common code to be reused by them.

Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>