perf: Revert back to pattern matching for ranges

May need to do more in depth checks and optimisations may need to be
removed if the core gleam compiler fixes the performance issues with
pattern matching + capture.
Possibly pattern matching may be better for smaller ranges but would
need to do benchmarks to find out what the cut-off point is.

src/gluri/internal/utils.gleam

@@ -372,17 +372,30 @@ fn unescape_percent(str: String) -> String {
 
 pub fn parse_hex_digit(str: String) -> Result(#(String, String), Nil) {
   case string.pop_grapheme(str) {
-    Ok(#(char, tail)) -> {
-      let assert [codepoint] = string.to_utf_codepoints(char)
-      let i = string.utf_codepoint_to_int(codepoint)
-      case i {
-        _ if i >= 0x30 && i <= 0x39 -> Ok(#(char, tail))
-        _ if i >= 0x41 && i <= 0x46 -> Ok(#(char, tail))
-        _ if i >= 0x61 && i <= 0x66 -> Ok(#(char, tail))
-        _ -> Error(Nil)
-      }
-    }
-    Error(_) -> Error(Nil)
+    Ok(#("0" as char, tail))
+    | Ok(#("1" as char, tail))
+    | Ok(#("2" as char, tail))
+    | Ok(#("3" as char, tail))
+    | Ok(#("4" as char, tail))
+    | Ok(#("5" as char, tail))
+    | Ok(#("6" as char, tail))
+    | Ok(#("7" as char, tail))
+    | Ok(#("8" as char, tail))
+    | Ok(#("9" as char, tail))
+    | Ok(#("a" as char, tail))
+    | Ok(#("b" as char, tail))
+    | Ok(#("c" as char, tail))
+    | Ok(#("d" as char, tail))
+    | Ok(#("e" as char, tail))
+    | Ok(#("f" as char, tail))
+    | Ok(#("A" as char, tail))
+    | Ok(#("B" as char, tail))
+    | Ok(#("C" as char, tail))
+    | Ok(#("D" as char, tail))
+    | Ok(#("E" as char, tail))
+    | Ok(#("F" as char, tail)) -> Ok(#(char, tail))
+
+    _ -> Error(Nil)
   }
 }