[RFC] Add new C.UTF-8 locale.

[Florian Weimer]

* Joseph Myers:

> On Mon, 29 Jun 2020, Florian Weimer via Libc-alpha wrote:
>
>> This refers to the entries for <UD800>, not the multibyte sequences.
>> I think we should aim for consistency between strcoll and wcscoll even
>> for invalid sequences.
>
> It's not clear what consistency means for byte sequences that cannot be 
> converted from narrow to wide characters or vice versa.  strcoll and 
> wcscoll have no way to report errors for such invalid byte sequences (so I 
> suppose it's implicitly undefined behavior).

I agree, but old glibc could converted between the two, so it's not a
stretch to talk of equivalence.

Thanks,
Florian