Replace langdetect lib by more updated fork

The language-detection library we use is https://code.google.com/p/language-detection/ version 1.1 from 2012. The project has stalled with no new development, not even in the [github repo](https://github.com/shuyo/language-detection) the original author put up.

Looks like the most promising fork is this one https://github.com/optimaize/language-detector/ which is also being selected by the Tika project to replace Tika's old detector.