In the type handlers for the full-text search types, we assume UTF8 in various places; this is very likely to produce bugs when full-text search is used against a database with another encoding.