non-unicode with ucwords. There is also an implementation for unicode ucwords
but that returns different results then strtotitle as it uppercases the
first character of every word, and doesn't *titlecase* a word. The test case
shows that.
Implement unicode_decode() and unicode_encode(), as described in README.UNICODE.
Still need to decide how to handle errors here, since there is no error return,
and a conversion error is not necessarily fatal.