Toward Embedding Bayesian Optimization in the Lab: Reasoning about Resource and Actions