However PyUnicode_AsUnicodeEscapeString() (which is used by
Objects/fileobject.c::file_repr()) still returns a str8 object.
Give unicode_repr() it's own implementation which returns a str8
object (it was formerly just calling unicodeescape_string()
which was used to implement PyUnicode_EncodeUnicodeEscape() too),
because once repr() is required to return unicode objects it
needs its own implementation anyway.