Calculates required A/B test sample sizes from power, significance, and MDE settings.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.