v0.3.1

Fixed Latin & Latin+CJK word segmentation:

• Latin: "abcd" is now correctly treated as one single token
• Latin+CJK: "package" was separated as "pack|a|ge", and now it's processed as one token