[COMMITTED PATCH] Separate Linuxisms from lowlevellock.h, make a generic one

[Roland McGrath]

> There is only zero semantic/minimal codegen effect for those machines
> currently affected by BZ#16892 - those being m68k, aarch64, arm and
> sh/sh4. So perhaps the thing to do is have those machines switch
> first, confirm the effect is as expected and then fix the bug before
> switching the others.

That seems reasonble.  But of course it is up to each machine maintainer.