RFC: Generate lib-names.h and stubs.h from ABI-LIST

[Roland McGrath]

> The only issue is ld=soname.  I can try to remove the soname part
> and use abi-ld-soname in CPU/Makefile instead.

This is going in the wrong direction.  It happens to be the case that
all the other soname strings match for x86_64 and x32, but that is not
something that the machinery should presume.  (In the abstract, there
is no good reason that x32 is using libc.so.6, for example--it's a new
space of sonames and so it could be libc.so.1 instead.  I imagine you
want to keep them matching just so as not to surprise people, and that
is reasonable enough.  But that's certainly not something that we
should be enforcing on people by having limiting assumptions in the
build machinery.)