The obesity paradox for outcomes in atrial fibrillation: Evidence from an exposure‐effect analysis of prospective studies