Fix inconsistent display width for emoji outside East Asian Wide ranges

[edugomz]

Fixes #4851

Problem

• display_width_of() only counted East Asian Wide/Fullwidth code points and two hand-picked emoji ranges (U+1F300-1F64F, U+1F900-1F9FF) as 2 columns wide
• Emoji outside those ranges (e.g. ❌ U+274C, ✅ U+2705, in the Dingbats block) were measured as 1 column, even though most terminals render them double-width
• Result: inconsistent {:^N} padding vs. CJK text

fmt::println("|{:^20}|", "❌");
fmt::println("|{:^20}|", "我");

|         ❌          |
|         我         |

Fix

• Replaced the boolean-expression range check with a sorted range table, looked up via binary search
• Table = original East Asian Wide/Fullwidth blocks + full Emoji_Presentation set from Unicode's emoji-data.txt (fetched from unicode.org, not hand-transcribed)

Testing

Added regression tests (test/format-test.cc) for:

• exact example from Inconsistent widths when UTF8 characters are used #4851
• multiple emoji back to back
• mixed emoji/CJK/ASCII content
• precision truncation at a display-width boundary
• alignment/fill variants
• newly-covered ranges (Transport, Symbols Extended-A)
• regional indicator (flag) pairs — documents current per-code-point behavior since fmt has no grapheme clustering

Full suite passes: 147/147 format-test, 21/21 ctest binaries.