Hardening follow-up: N+1 elimination, type validation, diversify fix

Additional fixes from parallel hardening streams:

- exercise/serializers: remove unused WorkoutEquipment import, add prefetch docs
- generator/serializers: N+1 fix in GeneratedWorkoutDetailSerializer (inline workout dict, prefetch-aware supersets)
- generator/services/plan_builder: eliminate redundant .save() after .create() via single create_kwargs dict
- generator/services/workout_generator: proper type-match validation for HIIT/cardio/core/flexibility; fix diversify type count to account for removed entry
- generator/views: request-level caching for get_registered_user helper; prefetch chain for accept_workout
- superset/serializers: guard against dangling FK in SupersetExerciseSerializer
- workout/helpers: use prefetched data instead of re-querying per superset

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

exercise/serializers.py

@@ -1,7 +1,6 @@
 from rest_framework import serializers
 from .models import *
 from muscle.models import ExerciseMuscle
-from equipment.models import WorkoutEquipment
 from equipment.serializers import WorkoutEquipmentSerializer
 
 class ExerciseMuscleSerializer(serializers.ModelSerializer):
@@ -26,13 +25,13 @@ class ExerciseSerializer(serializers.ModelSerializer):
         fields = '__all__'
 
     def get_muscles(self, obj):
-        # Use prefetched data if available, avoiding N+1 queries
-        if hasattr(obj, '_prefetched_objects_cache') and 'exercise_muscle_exercise' in obj._prefetched_objects_cache:
-            return [{'muscle': em.muscle_id, 'name': em.muscle.name} for em in obj.exercise_muscle_exercise.all()]
-        return list(obj.exercise_muscle_exercise.values('muscle', name=models.F('muscle__name')))
+        # Use prefetched related manager if available (avoids N+1 queries)
+        # Callers should use .prefetch_related('exercise_muscle_exercise__muscle')
+        objs = obj.exercise_muscle_exercise.all()
+        return ExerciseMuscleSerializer(objs, many=True).data
 
     def get_equipment(self, obj):
-        # Use prefetched data if available, avoiding N+1 queries
-        if hasattr(obj, '_prefetched_objects_cache') and 'workout_exercise_workout' in obj._prefetched_objects_cache:
-            return [{'equipment': we.equipment_id, 'name': we.equipment.name} for we in obj.workout_exercise_workout.all()]
-        return list(obj.workout_exercise_workout.values('equipment', name=models.F('equipment__name')))
+        # Use prefetched related manager if available (avoids N+1 queries)
+        # Callers should use .prefetch_related('workout_exercise_workout__equipment')
+        objs = obj.workout_exercise_workout.all()
+        return WorkoutEquipmentSerializer(objs, many=True).data