fix(dictionary): a Hanja reads 물 수, and each radical's form shows once

Korean names a Hanja by its meaning, then its sound; simplified Chinese
shares some radicals' forms with Japanese (麦), which showed twice, and
writes 黑 as itself. A test can read a language's strokes beforehand.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>

halkyon_learn_flutter/test/feature/kanji/model/radical_test.dart

@@ -49,6 +49,9 @@ void main() {
     expect(radicalFormsIn(food, 'zh-CN'), ['饣']);
     expect(radicalFormsIn(food, 'zh-TW'), ['飠']);
     expect(radicalsIn('zh-Hans')[148].variants, ['讠']);
+    // Each form once, never the radical itself.
+    expect(radicalFormsIn(radicals[198], 'zh'), ['麦']);
+    expect(radicalFormsIn(radicals[202], 'zh'), isEmpty);
     expect((hasRadicals('ko'), hasRadicals('el')), (true, false));
   });
 }

halkyon_learn_flutter/test/support/loaded_strokes.dart

@@ -1,18 +1,19 @@
 import 'package:flutter_test/flutter_test.dart';
 import 'package:halkyon_learn_flutter/src/feature/kanji/data/stroke_repository.dart';
 
-/// The bundled strokes of [characters], read already: reading an asset is
-/// real input, which a widget test's fake clock never lets finish, so it
-/// happens outside it and the app gets this repository
+/// The bundled strokes of [characters] in [language], read already:
+/// reading an asset is real input, which a widget test's fake clock never
+/// lets finish, so it happens outside it and the app gets this repository
 /// (`testDependencies(strokes: ...)`).
 Future<StrokeRepository> loadedStrokes(
   WidgetTester tester,
-  Iterable<String> characters,
-) async {
+  Iterable<String> characters, {
+  String? language,
+}) async {
   final strokes = StrokeRepository();
   await tester.runAsync(() async {
     for (final character in characters) {
-      await strokes.strokesOf(character);
+      await strokes.strokesOf(character, language: language);
     }
   });
   return strokes;