feat: add multi-country forecast evaluation pipeline

[SGHoekstra]

Summary

Adds utilities and example scripts for evaluating ABM forecast performance across countries and model variants (base, CANVAS, GrowthRate AR(1)).

• src/utils/bias_ttest.jl — one-sample HAC t-test (Newey-West SE) for forecast bias significance
• src/utils/analysis_utils.jl — forecast error aggregation, country discovery, CSV/LaTeX table writers
• src/utils/error_tables.jl — per-country RMSE/bias tables with AR and VAR benchmarks and Diebold-Mariano tests
• src/utils/save_all_predictions.jl — adds model_factory kwarg and run_variant_pipeline convenience wrapper
• examples/analysis/tabs/simulate_and_create_tables_multicountry.jl — runs the full pipeline for all calibrated countries
• examples/analysis/tabs/create_tables.jl — single-country table generation
• examples/analysis/tabs/aggregate_forecast_results.jl — aggregates per-country CSVs into cross-country summaries
• examples/analysis/figs/aggregate_forecast_results.jl — heatmaps of forecast performance across countries