Issue Title: Expose and manage multi-agent configurations in UI & backend

[Cataldir]

Background: Our essays and questions services already support multi-agent grading. To maximize flexibility, we need an admin interface and API to tailor which agents run and how they’re orchestrated.

Acceptance Criteria:

• Backend:
  • Update schemas.py (agents: List[Grader]) to accept dynamic configurations via API.
  • Support request flags (query params or body fields) to enable/disable individual agents at runtime.
• Frontend:
  • Implement an “Agent Config” UI under src/frontend/components/Configuration/Agents/.
  • List all available graders with controls to toggle, reorder, and adjust weights or prompts.
  • Persist user changes through the configuration service API.
• Shared Library:
  • Extract common orchestration logic into a shared module consumed by both /essays and /questions services.
• Testing:
  • Add unit tests for backend flags and frontend toggles to verify agent behavior changes.

[Cataldir]

Repository hygiene cleanup: this issue has been inactive since 2025-05 and has no active implementation branch mapped to it. Closing to keep open issues aligned to current work. If still needed, we can reopen and attach a new active branch.