Correct stale 1.56.14 main-residence evidence in the 1.56.16 UK register

[MaxGhenis]

The packaged 1.56.16 parity reference records
main_residence_value=0.674658
(packages/microcosm-build/src/microcosm/build/uk/efrs_parity_reference.json:313),
but the signed register and comparison ledger still quote 0.6761
(packages/microcosm-build/src/microcosm/build/uk/spine_swap_signed_differences.json:142;
experiments/686-uk-spine-comparison-ledger.md:184-190). The discrepancy
survives on current main. It does not alter direction or band classification,
but it disproves the stated completeness of the 1.56.16 cross-check.

Acceptance criteria:

• Correct the register and ledger from the pinned artifact.
• Add a machine check that every quoted incumbent share agrees with the
  packaged parity reference at the declared precision.

Verification and delivery record

• Required uv sync --all-packages --extra us was attempted first. A writable
  cache resolved the 123-package lock and package download then stopped on DNS.
• Focused Whole-spine parity for the UK: re-pinned incumbent, comparison instruments, signed register, and the E10 evidence (#686) #747 parity/register/payload/age/NHS tests exited 0. The production E6
  and adversarial parity probes produced the exact values above.
• Focused UK national calibration run-readiness: doctrine, live parity trio, US-format diagnostics, scorer, staging posture (#623) #743 seam/scorer/resolver/materializer tests and the exact expected-
  abort regression exited 0; fresh production-shaped/static probes produced
  every table verdict above.
• Tests used a sibling environment only after confirming its uv.lock SHA-256
  exactly matched this checkout, with this worktree's five source roots first
  on PYTHONPATH. Only focused tests/probes ran; no build or high-memory
  repository-wide job ran.
• Full reports are committed locally at review/pr-747-audit commit
  6b01dda54389621efe7fd8c90e281d7945223e7c and
  review/pr-743-audit commit
  6743d119d54f65cec3160bf7247d77a0d7a5f09b.
• Three authorized Git push attempts failed because the shell could not resolve
  github.com; the connector publication gate was not approved. Therefore the
  local review branches contain the required commits, but their remote refs
  were not updated. No unauthorized branch was pushed.

---

From the post-merge verification of #747 (round-1 review + fresh reproductions on merged main); full evidence: https://github.com/PolicyEngine/microcosm/blob/review/pr-747-audit/REVIEW-747.md

🤖 Filed by the microcosm owner session with Claude Code