[S390] Cleanup & optimize bitops.

The bitops header is now a bit shorter and easier to understand since
it uses less inline assembly. It requires some tricks to persuade the
compiler to generate decent code. The ffz/ffs functions now use the
_zb_findmap/_sb_findmap table as well.
With this cleanup the new bitops for ext4 can be implemented with a
few lines, instead of another large inline assembly.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>