feat: complete exam lifecycle — AI generation, submission, student session, and results

- Backend: AI generation fallbacks when OpenAI not configured, full exam
  submission saving all params (difficulty, rubric, entity, grading system,
  approval workflow) and creating linked question records per section
- Backend: new exam session controller with get_session, autosave, submit,
  status, and results endpoints; student attempt/answer/score models
- Backend: new controllers for entities, approval workflows, exam schedules
- Frontend: exam session split-layout with passage panel, question types
  (MCQ, T/F/NG, gap-fill, writing, speaking), timer, and review dialog
- Frontend: results page with percentage score, per-answer breakdown table
- Frontend: generation page dynamic dropdowns, full payload submission
- Frontend: updated types for ExamSessionSection, ExamQuestion options

Made-with: Cursor

backend/custom_addons/encoach_ai/services/openai_service.py

@@ -341,3 +341,6 @@ class OpenAIService:
         messages.append({"role": "user", "content": brief_text})
 
         return self.chat_json(messages, action=f"generate_{content_type}_dedup", max_tokens=4096)
+
+
+EncoachOpenAIService = OpenAIService