Respect dictionary encoding

[kolya-ay]

As soon as node-spellchecker tries to be as much system-native as possible, I think it makes sense to support non UTF-8 encoded dictionaries.

It seems that many distributions package their hunspell-* packages directly from libreoffice repo (so does Chrome with its own additions). Unfortunately, the encodings in the repo vary a lot. So to fully support system dictionaries we have to transcode strings. Probably, this way.

Related to #51, #54

[dmoonfire]

This should be resolved by #95.

[dmoonfire]

Since I haven't heard a response, I'm going to assume this is working. Feel free to reopen if it isn't.