468 use equivalent q for a b y nab in scaling method

[WesIngwersen]

cc: @MoLi7
Closes: #468

Summary

Extends the scaled-year A-matrix methodology to also produce a consistent scaled B matrix (E/x @ Vnorm) using equivalent summary-table scaling for both x and Vnorm.

New config flag: use_scaled_x_and_scaled_Vnorm_for_B

• Added use_scaled_x_and_scaled_Vnorm_for_B: bool to USAConfig with a validator that enforces scale_a_matrix_with_summary_tables or scale_a_matrix_with_ceda_method_as_fallback, and use_E_data_year_for_x_in_B, as prerequisites.
• Added config 2025_usa_cornerstone_full_model_scaling_for_A_and_B.yaml to exercise the full scaled A+B pipeline.

scale_cornerstone_x and scale_cornerstone_V (new)

• Exposed derive_summary_x_usa as a public function (was private).
• Added scale_cornerstone_x in cornerstone_year_scaling.py: scales detail industry gross output using summary x ratios, with optional dollar-year rebasing via the summary industry price ratio when adjust_summary_dollar_year_before_scaling is set.
• Added scale_cornerstone_V in cornerstone_year_scaling.py: scales the detail Make table using summary V block ratios, following the same pattern as scale_cornerstone_A but without the out-of-bounds clamping assertion (V row sums are not constrained to ≤ 1).

Rename: adjust_summary_A_and_q_dollar_year → adjust_summary_dollar_year_before_scaling

• Renamed the config field and all call sites across cornerstone_year_scaling.py, derived_cornerstone.py, YAML configs, and tests to better reflect that the flag now governs dollar-year alignment for A, q, V, and x.

Summary-level industry price index (new)

• Added _aggregate_industry_pi: x-weighted aggregation of Cornerstone-detail industry PI to BEA summary level using the bea_v2017_industry__bea_v2017_summary mapping.
• Added _get_summary_industry_price_index (cached): aggregates detail PI to summary using 2017 base-year Cornerstone x as weights (avoids a circular dependency through scale_cornerstone_x).
• Added get_summary_industry_price_ratio (cached): cross-year ratio of summary industry PI, used as the dollar-year deflator for adjust_summary_V_dollar_year and adjust_summary_x_dollar_year.
• Added adjust_summary_V_dollar_year: rebases summary V from one year's prices to another by dividing rows by the industry price ratio (row-axis only, consistent with inflate_cornerstone_V_with_industry_pi).
• Added adjust_summary_x_dollar_year: rebases summary x by dividing by the industry price ratio.

Scaled B matrix wiring (derive_cornerstone_Vnorm_scrap_corrected, derive_cornerstone_x_after_redefinition)

• When use_scaled_x_and_scaled_Vnorm_for_B is set, derive_cornerstone_Vnorm_scrap_corrected now uses scale_cornerstone_V for V and derive_cornerstone_x_after_redefinition for the scrap correction denominator. The x in the scrap correction uses the same summary industry output scaling as scale_cornerstone_A, keeping A and Vnorm consistent.
• derive_cornerstone_x_after_redefinition routes through scale_cornerstone_x(derive_cornerstone_x(), ...) when the flag is set, so derive_cornerstone_B_via_vnorm gets a detail x consistent with the A-matrix scaling denominators.
• derive_cornerstone_x now applies _distribute_waste_parent_x_using_v_row_shares so that waste-sector children receive proportional shares of parent gross output in all callers, including the PI aggregation weights in _get_summary_industry_price_index.

Analysis script

• Added bedrock/analysis/compare_x_approaches.py to compare derive_cornerstone_x_after_redefinition() (Approach A: summary x scaling) against compute_x(V=scale_cornerstone_V(...)) (Approach B: V row sums). Results show 401/405 sectors differ >1% (mean ~22%), driven by waste redistribution, confirming Approach A as the internally consistent choice for the B matrix.

Testing

Confirmed successful build of new model config 2025_usa_cornerstone_full_model_scaling_for_A_and_B