Extend error text span to whole code points

[MattX]

This fixes a panic when an error starts inside a Unicode code point. The range is extended to start (or end) at the beginning (or end) of the character inside which the byte offset is located.

Fixes #223. Unit tests provided in that issue by @Benjamin-L (lightly modified).

[zkat]

oh hey awesome! Thanks!

[Benjamin-L]

There's one more edge case here that might be worth fixing now or might be worth handling in a later PR:

    let src = "source\n  👼🏼text\n    here".to_string();
    let err = MyBad {
        src: NamedSource::new("bad_file.rs", src),
        highlight: (10, 0).into(),
    };

Produces

---- single_line_with_wide_char_unaligned_span_empty stdout ----
Error: oops::my::bad

  × oops!
   ╭─[bad_file.rs:2:4]
 1 │ source
 2 │   👼🏼text
   ·   ─▲
   ·    ╰── this bit here
 3 │     here
   ╰────
  help: try doing it better next time?

I would probably expect either

---- single_line_with_wide_char_unaligned_span_empty stdout ----
Error: oops::my::bad

  × oops!
   ╭─[bad_file.rs:2:3]
 1 │ source
 2 │   👼🏼text
   ·   ▲
   ·   ╰── this bit here
 3 │     here
   ╰────
  help: try doing it better next time?

or maybe

---- single_line_with_wide_char_unaligned_span_empty stdout ----
Error: oops::my::bad

  × oops!
   ╭─[bad_file.rs:2:3]
 1 │ source
 2 │   👼🏼text
   ·    ▲
   ·    ╰── this bit here
 3 │     here
   ╰────
  help: try doing it better next time?

[zkat]

@Benjamin-L oops, I didn't see your message before merging. Yeah, let's address that, too.