feat(RubricEvaluation): implement rubric auto-grading

[ncduy0303]

Description

Implement auto-grading service for rubric-based response questions. OpenAI LLM is used to evaluate student's answers and automatically assign suitable grades for different categories. A draft comment for overall feedback is also automatically generated where teachers can edit and publish later.

Changes made

• Implement rubric_auto_grading_service and rubric_llm_service
• Modify front end to dispay AI generated draft comments that teachers can modify and publish, and refresh views after the question is autograded.

Views

Teacher can see suggested grades and comments when opening student's submission. Pressing Re-evaluate Answer will rerun the autograding job. This button is in loading state if the answer is currently autograded, with the rubric panel being disabled too.

Comments can be modified and published in both the Submission Page and the Comment Center Page.

Notes

• Some UI enhancements on rubric panels were noted: rubric_grade being disabled dropdown, reevaluate button to moved down (and maybe changed name), 'moderation' category grade (putting -1, 10000, etc.), 'rubric explanation` overflow when explanation text is too long
• There are some outdated javascript files that might need to be refactored into typescript.

[cysjonathan]

• I am currently reusing the front end components ReevaluateButton that was used for autograding programming questions which called the back-end to auto grade the given answer.

OK

• Since it is autogradable, it is already triggered the moment student finalise their submission, so the grader will already see the rubric-based questions graded when they open the student's submission for the first time.

What is the behaviour if the grading is "in-progress"? i.e. when the instructor opens the submission before the grading completes.

• During manual mode, if the user clicks 'Evaluate Answers', all autogradable questions will be autograded again (which now included rubric-based question)

That is fine. I suspect we can also remove this 'Evaluate Answers' button since individual/single 'Re-evaluate Answer' is sufficient for manual triggering, and autograded answers will be autograded on finalise anyway.

• What should be the behaviour when the assessment setting is autograded? The attached screenshot is during manual mode.
  When assessment is autograded, as-per earlier point on autogradable question type triggers autograding on finalise, rubric-based question should autograde and assign the marks based on allocated explanations. The submission will then get a final score, similar to how we do autograding for programming questions (question-answer score will be calculated).

Autograding logic should go to app/services/course/assessment/answer/, this feature should not touch rag_wise, which is a separate service for responding to forum posts (and possibly comments in future). Can create a folder for rubric if needed.

Also add test cases once ready.

[adi-herwana-nus]

Looks mostly good, but there are some comments that need to be addressed. When making changes, remember to make them in the appropriate commits.

Also, there are endpoints called auto_feedback_count and publish_auto_feedback that allow instructor to publish all feedback from a given assessment at once, but currently these only affect programming file feedback by Codaveri.
A user would consider the draft comments generated as "automated feedback", so we should probably rename the button to something like "Publish Automated Programming Feedback" to disambiguate.