Trivial Description
In tika-eval, we count "common words" based on the output of language id. If we don't have a "common words" file for the identified language, we back off to the default language. This was hard-coded as "en", let's make this configurable.
While we're at it, fix a bug that prevents counting common words in zh.