Fix character counting in document statistics to use graphemes

- Add unicode-segmentation dependency for proper grapheme cluster support
- Replace chars() iteration with graphemes(true) for accurate character counting
- Fix counting of complex Unicode characters like emojis, combining characters, and multi-byte sequences
- Resolves TODO: 'do graphemes?' in document_statistics function

This change provides more accurate character counts for international text,
emojis with skin tones, combined characters, and other multi-codepoint graphemes.

Examples of improved accuracy:
- 👍🏾 now counts as 1 character instead of 2
- é (e + combining acute) counts as 1 character instead of 2
- 🧑‍💻 (person technologist) counts as 1 character instead of 4

Cargo.toml

@@ -24,6 +24,7 @@ syntect = "5.2.0"
 two-face = "0.4.3"
 # Internationalization
 icu = { version = "2.0.0", features = ["compiled_data"] }
+unicode-segmentation = "1.12"
 i18n-embed = { version = "0.16", features = [
     "fluent-system",
     "desktop-requester",