Fix interaction between debug presentation, precision, and width for strings

[localspook]

Consider our function that formats strings. When using debug presentation, it has 2 issues:

template <typename Char, typename OutputIt>
FMT_CONSTEXPR auto write(OutputIt out, basic_string_view<Char> s,
                         const format_specs& specs) -> OutputIt {
  auto data = s.data();
  auto size = s.size();
  if (specs.precision >= 0 && to_unsigned(specs.precision) < size)
    size = convert_precision_to_size(s, to_unsigned(specs.precision)); // Issue 1: we truncate the string
                                                                       // to fit within the precision...
  bool is_debug = specs.type() == presentation_type::debug;            //
  if (is_debug) {                                                      //
    auto buf = counting_buffer<Char>();                                //
    write_escaped_string(basic_appender<Char>(buf), s); // ...but don't account for the fact that it
    size = buf.count();                                 // can expand when escaped.
  } // ^^^^^^^^^^^^^^^^──────┐
                          // │ Issue 2: we reinterpret size in bytes as 
  size_t width = 0;       // │ display width. We can't do that, because
  if (specs.width != 0) { // │ debug presentation doesn't mean "just ASCII".
    width =     // vvvv──────┘
        is_debug ? size : compute_width(basic_string_view<Char>(data, size));
  }
  return write_padded<Char>(
      out, specs, size, width, [=](reserve_iterator<OutputIt> it) {
        return is_debug ? write_escaped_string(it, s)
                        : copy<Char>(data, data + size, it);
      });
}

Issue 1 manifests like this:

fmt::println("{:.2?}", "\n"); // Prints "\n", but should print "\

And Issue 2 like this:

fmt::println("{:*<5?}", "щ"); // Prints "щ"*, but should print "щ"** 
                              // (щ is two bytes, so {fmt} thinks it takes two terminal cells)

Godbolt link. This PR adds the logic to handle such cases.

[localspook]

Tangent, but: while working in this area of the code, I noticed we escape characters differently than the standard: we use \uXXXX and \UXXXXXXXX, while the standard uses \u{...}... is this intentional?