Refactor this 'working but unmaintainable' code into clean, testable, well-structured code. Explain every design decision. ```python def process(data, mode, flag1=False, flag2=True, output_type='json'): results = [] for item in data: if mode == 'fast': if item.get('type') == 'A': val = item['value'] * 1.1 if flag1 else item['value'] if val > 100: if flag2: results.append({'id': item['id'], 'val': val, 'status': 'high'}) else: results.append({'id': item['id'], 'val': val * 0.9, 'status': 'adjusted'}) else: results.append({'id': item['id'], 'val': val, 'status': 'normal'}) elif item.get('type') == 'B': val = item['value'] * 0.95 results.append({'id': item['id'], 'val': val, 'status': 'discounted'}) else: if flag1 and flag2: results.append({'id': item['id'], 'val': 0, 'status': 'skip'}) elif mode == 'careful': try: val = float(item.get('value', 0)) if val < 0: raise ValueError('negative') results.append({'id': item['id'], 'val': val, 'status': 'validated'}) except: results.append({'id': item['id'], 'val': 0, 'status': 'error'}) if output_type == 'json': import json; return json.dumps(results) elif output_type == 'csv': return '\n'.join([f"{r['id']},{r['val']},{r['status']}" for r in results]) return results ```