Add source-package facts for PE parity targets

[MaxGhenis]

Summary

• adds the source-package harness and consumer fact contract needed by Microplex target adapters
• adds PE parity source packages and raw/source manifests across US and UK publisher sources
• extends high-value packages for the current parity pass, including TANF state family counts, LIHEAP households served, SOI Historic Table 2 counts, state AGI amounts, ISC private-school pupil counts, and the HMT Budget 2025 salary-sacrifice contribution amount
• redacts the CMS page embedded public Mapbox token from the committed raw HTML while preserving the source-package parse path
• extends voa-council-tax-bands-2025 with 2,563 source-numeric local-authority band facts from CTSOP1.0, omitting suppressed cells rather than fabricating values
• extends PDF/HTML number parsing to preserve currency-prefixed compact amounts such as £32bn, and updates existing PDF guards to match the source text

Paired work

• paired Microplex US PR: Add Arch-backed PE target parity adapters microplex-us#9
• paired Microplex UK PR: Add Arch-backed UK target provider microplex-uk#1
• US source-backed coverage result on the paired Microplex branch: 172 / 172 selected PE native broad source-backed cells covered
• UK checklist result in Arch: 27 / 28 items checked; remaining blocker is the DWP Scotland UC child-under-1 Stat-Xplore export/API

Validation

• uv run arch build-suite hmt-budget-policy-costings-2025-salary-sacrifice --year 2025 --out /tmp/arch-hmt-budget-policy-costings-2025-salary-sacrifice --replace -> valid, 1 fact
• uv run pytest tests/test_arch_source_package.py -q -k "hmt_budget_salary_sacrifice or hmt_budget_policy_costings or hmrc_salary_sacrifice_reform or hmrc_salary_sacrifice_relief" -> 6 passed
• uv run pytest tests/test_arch_source_package.py tests/test_arch_suite.py -q -k "soi_table_1_2_facts or isc_census_pupil_count or voa_council_tax_band or cms_medicare_trustees_part_b_premium or treasury_eitc_outlay or new_us_source_counts or build_source_suite_supports_soi_table_1_4" -> 59 passed
• uv run pytest tests/test_arch_source_package.py -q -> 227 passed
• uv run pytest tests/test_arch_suite.py -q -> 13 passed
• uv run ruff check arch/sources/cells.py arch/source_package.py arch/suite.py tests/test_arch_source_package.py tests/test_arch_suite.py -> passed
• git diff --check -> passed
• uv run pytest tests/test_arch_source_package.py -q -k "isc_census or isc_census_pupil" -> 2 passed
• uv run arch validate-package packages/isc/census_2024 --year 2024 -> valid, 1 source record / 1 measure
• uv run arch build-suite packages/isc/census_2024 --year 2024 --out /tmp/arch-suite-isc-census-2024-current --replace -> valid, 1 consumer fact with concept alignment
• uv run pytest tests/test_arch_source_package.py -q -k "hhs_acf_tanf_caseload or hhs_acf_liheap_profile or soi_historic_table_2_state_agi or new_us_source_counts" -> 55 passed
• uv run pytest tests/test_arch_source_package.py::test_source_package_alias_builds_cms_medicare_state_payment_facts -q -> 1 passed
• uv run arch build-suite soi-historic-table-2-state-agi-2022 --year 2022 --out /tmp/arch-suite-soi-historic-table-2-state-agi-2022 --replace -> valid, 918 facts
• uv run arch build-suite hhs-acf-liheap-fy2024-national-profile --year 2024 --out /tmp/arch-suite-hhs-acf-liheap-fy2024-national-profile --replace -> valid, 1 fact
• uv run arch build-suite hhs-acf-tanf-caseload-2024 --year 2024 --out /tmp/arch-suite-hhs-acf-tanf-caseload-2024 --replace -> valid, 58 facts
• uv run arch build-suite soi-historic-table-2 --year 2022 --out /tmp/arch-suite-soi-historic-table-2-2022 --replace -> valid, 143 facts
• uv run arch validate-package voa-council-tax-bands-2025 --year 2025 -> valid, 2,653 source records / 2,653 consumer facts after the local-authority extension
• uv run arch build-suite voa-council-tax-bands-2025 --year 2025 --out /tmp/arch-suite-voa-council-tax-bands-2025-current --replace -> valid, 2,653 consumer facts, 0 agent acceptance errors
• uv run arch validate-package scotgov-council-tax-bands-2025 --year 2025 -> valid, 9 source records
• uv run arch build-suite scotgov-council-tax-bands-2025 --year 2025 --out /tmp/arch-suite-scotgov-council-tax-bands-2025-current --replace -> valid, 9 consumer facts, 0 agent acceptance errors

[MaxGhenis]

Cycle pass update (2026-05-27):\n\n- Merged the stacked storage-config PR into this branch earlier.\n- Removed stale Cosilico package/env/schema references from this draft branch.\n- Redacted token-like values embedded in mirrored HTML source files and updated the affected source manifests/checksums.\n- Local validation passed: ruff check README.md db/supabase_client.py tests/test_supabase_client.py docs/architecture.md docs/repository-model.md, git diff --check, focused source-package redaction checks (16 passed, 1 skipped), and the broader source-package/suite/Supabase set (250 passed, 1 skipped).\n\nI am keeping this PR as draft for now. The remaining blocker is reviewability/repo hygiene: it still bundles core harness/schema work with about 155 MB of raw publisher files and generated-scale source-package YAML. Recommended next step is to split this into: (1) core harness/schema/tests/docs, (2) compact source package specs, and (3) raw artifact/R2 registration or fixtures, keeping large raw bytes/generated outputs out of the main merge path where possible.

[MaxGhenis]

Closing this omnibus draft as superseded by the split Arch source-package PRs now merged into main. For the Microplex US/eCPS replacement path, the relevant source-package and adapter work has landed through the focused Arch PR sequence (#7 through #48) and paired Microplex PRs, with current Microplex source-backed broad coverage verified at 183/183. Keeping this draft open would be misleading because it still bundles old core harness work with large raw publisher files and unrelated UK/source-package experiments. Any remaining UK/eFRS or non-US source packages should come back as focused PRs with their own package validation/build-suite evidence.