Bugfix grapheme and East-Asian support by wcwidth>=0.5.1 integration

[jquast]

• Bugfix In _CustomTextWrap._handle_long_word(), on master branch or this branch when optional co-dependency wcwidth<0.3.0:

  • east-asian characters were measured using len() but should use width-aware self._len().
  • Tests are provided to exercise this.

• Improve word-wrapping with unicode, starting wcwidth version 0.3.0, the wrap(), and width() functions also parse sequences pretty much the same, even a bit better by "tracking" more kinds of SGR sequences and "reducing" them a bit better on line breaks.

  • optional wcwidth>=0.5.1 handles grapheme boundaries better for many kinds of languages, or just flags and emojis and significantly improved performance over 0.2.14.

note

• Because almost all users will not know to install the optional "wide" version, they may also have wcwidth at an older version installed by chance. hasattr() must be used for some defense of this circumstance.
  • a special test fixture is used to exercise both possibilities even when latest wcwidth is installed by a mock "wcwidth" module that pretends to be without wrap().

[jquast]

functools.partial() was just used to help make this change clear and concise, not needed but doesn't hurt and already imported and used somewhere else

[astanin]

I would be helpful to know what is the original issue this PR is trying to solve.
Input, output, expected output.

Please add or reference an issue.

[jquast]

That's ok, I thought the automatic tests were self-evident. the issue is yours to solve, I don't have any more time to volunteer to your project. best wishes!!

[astanin]

OK, thanks!