Potential Pilot Problems: Treatment Spillovers in Financial Regulatory Experiments