Improve autowidth on Unicode character

[ickc]

Currently, autowidth only calculate width by no. of characters.

I recall @jgm does something smart in pandoc to account for the width of unicode character. e.g. Chinese character is considered twice as long as an ASCII.

Considerations:

• what if its other unicode character?
• do we need to match pandoc's behavior exactly? Note that this should not affect idempotency (because once the numerical value of width are set, it's done).
• are there any better width prediction algorithm?
• what does this mean to performance?

[ickc]

lack of interested and not quite simple to do