[Rishiii57]

Hey team,

This fixes #2414 so contests that haven't been rated yet aren't prematurely rated when re-rating after a disqualification.

What was happening

When a user is disqualified from a past contest, contest.rate() runs to recalculate ratings for that contest and every later contest up to the current time. The query previously grabbed all contests where is_rated=True in (self.end_time, self._now) without checking if those contests had ever been rated yet.

If a contest finished recently but organizers were still holding off on rating it (for example, while reviewing plagiarism or waiting on editorial releases), re-rating an older contest would prematurely publish ratings for that newer contest.

What changed

In judge/models/contest.py, I updated Contest.rate():

1. Filter contests to only include self (to ensure the target contest is always rated, even when rating it for the first time) or contests that already have ratings (ratings__isnull=False):

   contests = list(Contest.objects.filter(
       Q(id=self.id) | Q(ratings__isnull=False),
       is_rated=True, end_time__range=(self.end_time, self._now),
   ).order_by('end_time').distinct())

2. Evaluate contests into a list before deleting old ratings, ensuring Rating.objects.filter(contest__in=contests).delete() only targets the contests that are actually being re-rated, without wiping out the query's ability to detect which contests were already rated.
3. Re-rate only the filtered list of contests. Any contest in the time window that has never had ratings computed remains unrated until its organizers explicitly trigger its rating.

Testing

• Verified with a standalone reproducer (reproduce_dmoj_2414.py) confirming that an unrated contest in the time window is prematurely rated on unpatched code and cleanly skipped on patched code.
• Added ContestRateTestCase in judge/models/tests/test_contest.py testing the Q filter expression and verifying that only self and already-rated contests are deleted and passed to rate_contest().
• Passed all unit tests via manage.py test with 0 errors.
• Passed flake8 with 0 warnings.