Related to #62.

This should primarily affect emoji as most other multi-chararcter graphemes tend to involve zero width trailing characters, so the overall grapheme width is given by the base code point.

Ideally this would be resolved by support directly in R.