refactor(exercises): standardize answer language handling across exercise scripts

- Introduced a mechanism to infer answer language based on question phrasing in multiple exercise scripts, enhancing consistency in exercise data.
- Updated question formats to clarify the intent of exercises, improving user understanding and engagement.
- Streamlined the code for better maintainability and clarity in exercise generation processes.

backend/scripts/update-feelings-affection-exercises.js

@@ -349,7 +349,8 @@ function createFeelingsAffectionExercises(nativeLanguageName) {
       instruction: 'Übersetze den Bisaya-Satz ins ' + nativeLanguageName,
       questionData: JSON.stringify({
         type: 'multiple_choice',
-        question: `Wie sagt man "${conv.bisaya}" auf ${nativeLanguageName}?`,
+        answerLanguage: 'native',
+        question: `Was bedeutet "${conv.bisaya}"?`,
         options: options
       }),
       answerData: JSON.stringify({
@@ -379,6 +380,7 @@ function createFeelingsAffectionExercises(nativeLanguageName) {
         instruction: 'Was bedeutet dieser Bisaya-Satz?',
         questionData: JSON.stringify({
           type: 'multiple_choice',
+          answerLanguage: 'native',
           question: `Was bedeutet "${conv.bisaya}"?`,
           options: options
         }),