Encoding API

[irinakhismatullina]

Encoding happens exactly the same way it is done in YouTokenToMe.

• Input - space-separated words. No normalization is performed
• Each word is tokenized separately, special bow (or space) token is added to the beginning of each word.
• Left-first order of tokenization (aaa -> aa a and not a aa).
• Both numerical encoding and tokens are returned at once. This may be changed if needed.

Closes #4

[vmarkovtsev]

Good job @irinakhismatullina ! I like where we are heading towards.