Fix POSIX::strxfrm()

[khwilliamson]

This function takes an SV containing a PV. The encoding of that PV is based on the locale of the LC_CTYPE locale. It really doesn't make sense to collate based off of the sequencing of a different locale, which prior to this commit it would do if the LC_COLLATION locale were different.

[bram-perl]

Is it possible to add test case for this?

[khwilliamson]

I don't see a good way for such a test case. The results of strxfrm() are opaque, varying by platform and potentially by release. And finding a locale that was sufficiently different from another locale would be hard to do portably. The function now is just a wrapper around the function that is used all the time by perl code, and so is tied to that working or not. That will be tested more in a later commit that runs a bunch of locales in parallel threads

[bram-perl]

If a test case is not possible, is an example - that could be included in the commit message - possible? That is: can you add a - non portable - example in the commit message where the output of POSIX::strxfrm is different before/after the commit?

[khwilliamson]

@bram-perl the commit message has now been rewritten and includes an example from glibc.