Only wrap on actual spaces

Trying to match on Unicode character classes to determine where to
wrap lines turns out to be a fragile and blacklist-heavy exercise.
Instead, just wrap on actual (ASCII \x20) space characters. This
fixes a bug where the escape characters starting ANSI sequences were
sometimes matching as a wrap point garbling the output, but also
simplifies the code by no longer needing to exclude carriage returns
in a CR+LF line ending sequence. Further it solves a couple TODO
reminders to stop wrapping on non-breaking (Unicode \xa0) space
characters.