unicode_width() now correctly returns 2 for the CJK and other code points that are typically displayed in 2 columns with non-proportional fonts.