feat(moderation): implement semantic activity logs UI and scalable architecture

[mnietona]

Description

This PR builds upon the Moderation Dashboard to introduce a clean and scalable Activity Logs system.

Instead of dealing with raw database field names and complex if/else logic in the HTML, I refactored the ModerationLog model using Object-Oriented Programming (OOP) principles. The model now acts as a "translator" that formats its own data, making future additions trivial.

Key implementations:

1. Scalable OOP Architecture: Added @property methods (action_text, action_color, target_text) to ModerationLog. If we add new actions in the future (e.g., hiding a document, archiving a course), we just need to update the model. The UI will adapt automatically without touching any HTML.
2. Semantic Tracking: Refactored views.py to stop logging raw, noisy database fields. We now log explicit semantic actions (action_accepter, action_rejeter) for cleaner auditing, and exclude old technical noise from queries.
3. Scrollable Dashboard UI: Added a sticky-header, scrollable activity table to the main moderation dashboard so moderators can see recent actions at a glance.
4. Django Admin Cleanup: Removed the now-redundant RepresentativeRequest from the Django admin and applied the semantic properties to the ModerationLog admin view for a perfectly clean backend.

📸 Screenshots

1. Scrollable UI on the Dashboard

2. Cleaned up Django Admin Interface

⚠️ Note for review

This PR is stacked on top of the feat/moderation-dashboard branch. It should be reviewed and merged after the initial moderation hub PR.