P2C10: reproduce measured correction over BLS - #13

What changed

• add a hermetic BLS JOLTS release/erratum pair as distinct immutable World Observations
• compare an exact corrected-statement artifact with a stale-form control under identical source coverage, correction linkage, reviewed Action topology, and matched conditions
• record independently authenticated reviews and exact Core Outcomes, then reuse the unchanged domain-neutral measured-impact service
• freeze the work packet, point-in-time audit, negative controls, and restrained README/roadmap candidate state
• add an optional P2C9 state handoff so the stacked journey remains append-only

Why

P2C9 still exercised only the Federal Register source family. P2C10 falsifies source-policy coupling against materially different real BLS public errata without widening Core, Intelligence, the inert Domain Pack, or either connector.

Under the frozen World rule, treatment scores 1.0 and the stale-form control scores 0.0 across two matched reviewed workflows. Intelligence classifies the bounded effect as useful. The resulting promote proposal is non-effective, non-selectable, unapplied, and requires separate human review.

Verification

• measured P2C3–P2C10 stack: 30 passed
• complete stacked World suite: 113 passed
• Federal Register connector: 26 passed
• release contract: 7 passed
• public ace-core==0.5.0 boundary: 83 passed, 30 explicit candidate skips
• scoped Ruff check/format: pass
• unchanged 0.9.0 sdist/wheel: 45 inert JSON/metadata entries, no Python or entry points, ace-core>=0.5.0,<0.6
• git diff --check: pass

Repository-wide Ruff remains inherited and unchanged from the P2C9 parent: 14 lint findings and 12 format targets.

Claim boundary

The historical wrong form is derived from BLS's explicit missing-minus-sign erratum because the current archived release is already corrected. This is one recorded-replay correction, not original-response-byte preservation, live monitoring, JOLTS statistical validation, population correction performance, causality, general source independence, or human benefit. It does not release World Intelligence, close ACE Core #38, pass SI4, or complete ACE 0.6.0.

Core issue #49 F1, F3, and F5 still need explicit 0.6 release-owner disposition; this packet does not implement or re-date them.

Stacked on World P2C9 (#12) and the ACE Core measured-impact candidates (#88–#90).