fixes for utf-8 to/from unicode conversion

The functions utf8_to unicode and unicode_to_utf8 didn't work correctly
with some 3 and 4 byte strings. This exchanges those functions against
two more sophisticated ones. It also adds unit testing for them.

darcs-hash:20060221212605-7ad00-7bfefe8c9615d5a7f3b33c279ce79d4200d4778c.gz