[codex] Add explicit PE SOI path override for rebuild CLI

[anth-volk]

Fixes #147

Summary

• Keeps the existing repo-based SOI route unchanged: policyengine_us_data/storage/soi.csv under --policyengine-us-data-repo.
• Adds explicit soi_path plumbing through the default PE-US rebuild source-provider factory and checkpoint CLI (--soi-path).
• Adds regression coverage that the source-provider factory forwards soi_path to the PUF provider.

Root Cause

The PUF PE-SOI uprating path expects policyengine_us_data/storage/soi.csv when only --policyengine-us-data-repo is provided. In the affected fresh-checkout run, that file was absent while the tracked calibration target file existed elsewhere. Rather than guessing alternate storage paths, callers can now supply the intended SOI file explicitly with --soi-path.

SOI Route Check

• microplex-us default repo route: <policyengine-us-data repo>/policyengine_us_data/storage/soi.csv.
• Current PE-US-data download_prerequisites.py does not list soi.csv; it downloads PUF/demographic/NP and geography prerequisites, not this SOI artifact.
• In the local PE-US-data checkout, policyengine_us_data/storage/soi.csv exists but is not tracked; policyengine_us_data/storage/calibration_targets/soi_targets.csv is tracked. Use --soi-path for that alternate file when needed.

Validation

• uv run --no-sync ruff check src/microplex_us/data_sources/puf.py src/microplex_us/pipelines/pe_us_data_rebuild.py src/microplex_us/pipelines/pe_us_data_rebuild_checkpoint.py tests/test_puf_source_provider.py tests/pipelines/test_pe_us_data_rebuild_checkpoint.py
• uv run --no-sync python -m py_compile src/microplex_us/data_sources/puf.py src/microplex_us/pipelines/pe_us_data_rebuild.py src/microplex_us/pipelines/pe_us_data_rebuild_checkpoint.py tests/test_puf_source_provider.py tests/pipelines/test_pe_us_data_rebuild_checkpoint.py

Notes

• I did not run the new pytest cases in this branch because main still has the no-torch core import issue addressed separately by [codex] Split PE rebuild CLI parsing from runtime imports #154; test-module import stalls in this no-torch environment before reaching the new PE-US rebuild smoke expects missing policyengine_us_data/storage/soi.csv #147 case.