Some of postgresql's string functions behave differently
depending on the locale the cluster was created with. With the locales
'C' or 'en_US.utf8', lower() does not lowercase non-ascii characters
such as Ü (umlaut-u).
If we have users from all over the world storing data in our
database with all sorts of languages, which locale should we use to
ensure these string functions behave as they expect?