All lowercase letters are now properly converted to uppercase letters (and vice
versa), via functions `towupper` and `towlower`.
Functions `string_tolower`, `string_toupper` and `utf8_charcasecmp` have been
optimized to be faster when there are ASCII chars (< 128); functions are about
25-40% faster with mixed chars (both ASCII and multi-bytes).
Function `utf8_wide_char` has been removed, `utf8_char_int` can be used
instead.
Now the function utf8_char_size_screen can return -1 when the char is not
printable.
It has a specific behavior for some chars:
- U+0009: value of option weechat.look.tab_width
- U+0001 to U+001F (except U+0009): 1
- U+00AD (soft hyphen): -1
- U+200B (zero width space): -1
When there non printable chars, the return of the function was 1.
For example utf8_strlen_screen("abc\x01") now returns 4 instead of 1.
In addition the function has been optimized to not use the `mbstowcs` function
which is slow; result is up to 15% faster.
Tests for size on screen are made on U+2EE9 (CJK Radical Simplified Yellow)
instead of U+24B62 (CJK Unified Ideograph-24B62) which returns a length of 1 on
FreeBSD and 2 on Linux.
Sébastien Helleu