fix(utils): surprising behavior in truncate_str when tail is larger than width

[remi-dupre]

Hello!

While looking again at #206 I bumped into a few issues in current implementation for truncate_str.

1. It may truncate if not required if measure_width(s) + measure_width(tail) > width : truncate_str("foo bar", 7, "!") == "foo ba!". I don't think this is the expected behavior.
2. Using a tail longer than text width leads to overflowing substractions, which gives "undefined behavior" (ish) :

// In debug mode, panics:
truncate_str("foo bar", 2, "!!!")  # panics
// In release mode, weird result:
truncate_str("foo bar", 2, "!!!") == "foo bar"

[djc]

Makes sense, thanks!