These tests require the test database to be in UTF8 encoding.  Until
there is a better solution, take them out of the default test set and
treat them like the existing collate.linux.utf8 test, meaning it has to
be selected manually.