Synthetic difference-in-differences shutdown attribution

For a country + date, runs synthetic-control SDiD attribution. Returns the counterfactual block_rate (from weighted stable-democracy donors), causal_effect estimate, permutation p-value, donor weights, and any nearby political events. Method: Arkhangelsky et al. arXiv:1812.09970 + Internet Society NetLoss (ACM JCSS 2024). Honest gating: low_confidence=true if pre_rmse > 0.10 or sample too small.