detectlanguages should detect languages in FoLiA

[kosloot]

The --detectlanguages option is confusing:
On plain text it means: Detect the language, tokenize according to that language and assign it to the FoLiA output.
On FoLiA input it means: check the language tag of elements and when it is in the provided list, tokenize it, according to the language.

I think that in FoLiA input it should be possible to really detect the language too.
This probably only will work correctly on input documents without any language info, but still that is useful.