Accuracy of claims‐based algorithms for epilepsy research: Revealing the unseen performance of claims‐based studies