Add partial NZ 2025 population source facts

[MaxGhenis]

Summary

Partial progress on #176's demographic source packages. This does not close #176 or complete the NZ age/sex target surface.

• Add 85 source facts from the pinned Stats NZ 30 June 2025 provisional subnational workbook: 16 regions plus the independently published NZ total, each across all ages and four broad age bands (0–14, 15–39, 40–64, 65+).
• Preserve the 30 June 2025 reference date, 29 October 2025 publication, 2023 census-controlled ERP definition, and 1 January 2025 REGC boundaries. National totals come directly from source cells, never regional sums.
• Pin the original publisher workbook, official REGC code/name attributes, and the national release landing page as evidence only. All three raw objects were uploaded under raw/nz/stats_nz/... and downloaded again with matching SHA-256 hashes and byte counts.
• Keep the wave-1 package-completion boxes unchecked. The detailed regional five-year-age × sex cut is not in the workbook; the POPES_SUB_004 structure request returned HTTP 401. The national June 2025 release points to Infoshare and provides no single-year-age workbook download; no such source package or invented age values are added.
• Add exact-cell, boundary/code, point-in-time vintage, age-universe, source-drift, and consumer-contract regression gates. Source artifacts are marked binary in Git attributes to preserve publisher line endings and spacing.

Verification

• uv run chronicle validate-package stats-nz-subnational-population-estimates-2025 --year 2025: valid, no package warnings.
• Bounded build-suite: valid; 85/85 lineaged facts, 2,265 parsed cells, 187 constraints, zero acceptance errors.
• uv run pytest -q tests/test_chronicle_nz_population.py tests/test_chronicle_artifacts.py tests/test_source_package_alias_drift.py --disable-warnings: 62 passed (12 existing dependency deprecation warnings).
• Scoped Ruff checks/formatting and staged git diff --check: pass.
• Explicit single-package bundle regression: pass. Whole-bundle golden counts reflect the measured 85-fact addition to base 10597ae; the full multi-country build was not run locally, to respect the battery constraint. CI must verify it before landing.

Review boundary

All concept mappings use source_label. The source build reports the expected warning that Axiom CLI canonical-concept validation did not run. This PR does not claim canonical certification, population calibration, target activation, uprating, reconciliation, or a usable microsimulation dataset.

Independent ledger-source-fidelity and ledger-boundary review is requested before landing. The detailed source gaps and original-vintage caveats are in docs/pe-nz-population-source-notes.md.