Counterfactual reasoning for LegionIO — imagining alternative outcomes, computing regret/relief, and extracting lessons from what-if scenarios