Update ML candidate sheet for data scientists before retrain

Your candidate records land out of sync between the ATS and model sheet, causing stale features and model drift before retrains. Ensure data scientists receive retrain-ready inputs so the next model run stays accurate.