PATCH: eliminate some endian #ifs

[Greg McGary]

Greg McGary <greg@mcgary.org> writes:

> Ulrich Drepper <drepper@redhat.com> writes:
> 
> > > No build-shared & build-static failures or binary differences for i686.
> 
> Same deal.
> 
> I refined this a bit further: __LONG_LONG_SPLIT accepts a single
> 64-bit quantity, breaks it into two 32-bit args, and passes the pieces
> according to __BYTE_ORDER.

FYI, When all the dust settled, I got binary differences with this
one, so I'll just drop it.  I'm going to check in my first patch which
had no binary differences.