A penalization approach to random‐effects meta‐analysis