Correct the handling of 0-termination of PyUnicode_AsWideChar()

and its usage in PyLocale_strcoll().

Clarify the documentation on this.

Thanks to Andreas Degert for pointing this out.