Move eCPS calibration base year from 2024 to 2025

[MaxGhenis]

Problem

The current national dataset and calibration pipeline are still fundamentally wired to 2024, but ACA behavior has already diverged materially in 2025. PR #622 adds a narrow 2025 post-calibration ACA enrollment override, which is useful as a bridge, but it knowingly produces a published dataset whose ACA enrollment no longer comes from the same calibration path that generated the rest of the weights.

The cleaner fix is to move the eCPS calibration base year to 2025 rather than accumulating one-off post-calibration overrides.

Goal

Produce a real 2025-calibrated eCPS dataset and supporting calibration artifacts, so that published 2025 outputs are coherent with the target system and do not rely on the ACA-specific override.

Scope

• move the calibration period for the production eCPS build to 2025
• update target ingestion and uprating so policy_data.db supports a coherent 2025 calibration run
• add year-aware dataset and upload plumbing for 2025 calibration artifacts
• remove the 2025 ACA post-calibration override once the 2025 path is ready

Proposed breakdown

1. Add 2025 target coverage and uprating rules

• add explicit 2025 ACA and Medicaid target coverage
• document which 2025 targets are direct admin/statistical values versus uprated inputs
• make sure legacy target readers do not silently keep pulling 2024-only ACA/Medicaid files during a 2025 build
• align with the newer database-backed calibration path where possible

2. Parameterize calibration artifacts by year

• add EnhancedCPS_2025 and any matching reweighted/source-imputed dataset classes needed for a production build
• make upload, validation, and Modal plumbing year-aware instead of hardcoding 2024 filenames
• keep 2024 and 2025 artifacts side by side during the migration

3. Switch the production default to 2025 and remove the override

• run the production calibration flow with time_period=2025
• update tests, staging validation, and docs
• remove or disable the ACA-specific post-calibration override introduced as the temporary bridge

Acceptance criteria

• published 2025 datasets do not rely on a post-calibration ACA enrollment override
• the production calibration path can build and validate a 2025-calibrated dataset end to end
• target ingestion is explicit about which 2025 values are observed versus uprated
• repository docs, Make targets, and upload scripts no longer assume 2024-only production artifacts

Related PRs / context

• PR Add 2025 post-calibration ACA enrollment override #622: temporary 2025 ACA enrollment override
• PR Update SOI-backed calibration targets through TY2023 #660: newer SOI-backed calibration target updates
• issue Hardcoded 2024 dollar values in etl_national_targets.py will break for future years #515: hardcoded 2024 values in national target ETL