Long‐term studies and reproducibility: Lessons from whole‐lake experiments