Correctly handle utf-8 string arguments in alignment

[vitaut]

As pointed out by Артём Голубихин in the comments to https://www.youtube.com/watch?v=ptba_AqFYCM:

Alignment is the problem, because utf-8 symbols are treated as a couple of chars, so, for example, u8"{:>2}"_format(u8"я") is just "я" instead of " я". And, of course, not just single unicode codepoints should be considered.

[vitaut]

Fixed in 3832524 for UTF-8 strings.

[gulrak]

Actually this still seems to be wrong for some codepoints, as e.g. u8"작" is now handled as having a width of 1 (as it is one codepoint) but actually when printing it on a terminal, it is two characters wide.

For systems supporting it, int wcwidth(wchar_t c); from <wchar.h> will give the visual width in character columns if current locale supports unicode, otherwise something along the lines of https://www.cl.cam.ac.uk/~mgk25/ucs/wcwidth.c might be needed.

[vitaut]

@gulrak, you are right, some characters have double display width. Thanks for the link! I plan to improve handling of width and the ideas from the wcwidth implementation are very useful.