Magyarlanc tokenizer to work (almost) like QunToken

[DavidNemeskey]

@sassbalint And there you have your pull request. :smile: These changes address all of the differences I have listed in the table, with the following caveats:

• kind is provided on a best-effort basis (a token is punctuation if it consists only of punctuation characters)
• the last whitespace in a batch is not added as a token. QunToken adds it, but in ML the text is trimmed, and I didn't want to change that

I also did a few cosmetic changes in the QunToken wrapper.

I have not included the jar (Lang_Hungarian/hungarian.jar), in case I have to make changes to this pull request. If you think it can be merged, I will compile and add the jar then.

[sassbalint]

Wow! :) Thank you. I'll look at it.

[sassbalint]

@DavidNemeskey Thank you for your contribution. :)