Time-sensitive RCTs in behavioral public policy: a pragmatic framework using sequence methods, personalization, and reinforcement learning | Synapse