fix: inconsistent loading of discussion thread response

[naincy128]

Description

Optimizes discussion comment loading by replacing in-memory sorting/pagination with DB-level operations and removing dead code that fired unnecessary queries per comment.

Root Cause

Comment.get_list() loaded all matching comments into memory using Python sorted(), then sliced — with no select_related, causing ~4 FK queries per comment in to_dict().

Additionally, CommentSerializer.to_representation() re-fetched the full thread and comment from the DB for every comment (~15 queries each) to perform an endorsement check that never executed — the condition "endorsement" not in comment_from_db was always False since to_dict() always includes the key.

For 10 comments, this resulted in ~258 queries and ~30s load times.

Changes

• Replaced Python sorted() with DB-level order_by() using F('sort_key').asc(nulls_last=True)
• Replaced Python slicing with QuerySet slicing (DB LIMIT/OFFSET)
• Added select_related('author', 'parent', 'comment_thread', 'deleted_by') to eliminate FK N+1
• Removed dead get_thread() and get_comment() calls from to_representation()

Impact

~258 queries → ~59 queries (77% reduction), load time ~30s → ~8-10s

Linked PRs

edx-platform
frontend-app-discussions

Ticket

LP-529