Optimize backend database queries in grading workflow

[AishaA-cpu]

Problem

Multiple redundant or sequential database queries in grading.py slow down the grading pipeline. Opportunities exist to use advanced SQLAlchemy loading techniques (joinedload, selectinload) for batch-fetching related entities.

Proposal

• Refactor grading workflow to prefetch assignments, IO test cases, unit tests, static rules, etc. for each submission
• Use joinedload/selectinload for one-to-many or many-to-one relationships
• Add appropriate indexes if any filter fields show up in performance analysis

Rationale

• Reduces DB roundtrips per grading task
• Lowers contention as grading task throughput increases
• Simplifies transactional logic and preloading

Acceptance Criteria

• Grading a submission involves minimal DB queries for related data
• Queries use batch/optimized loading, evidenced via logs or analysis
• No unrelated performance regressions

---

Labels: backend, enhancement, performance