CK-07R1: correct lifecycle preparation scale

[douglasmonsky]

Summary

• linearize lifecycle folding by grouping new transitions once per entity
• add publication-valid standard and production synthetic scale qualification
• bind frozen CK-08R0 budgets, producer digests, writer receipts, and database postconditions

Validation

• focused lifecycle, documentation authority, and scope tests: 29 passed
• five-sample replay: standard max 42.980 ms; production max 20,666.089 ms
• ordinary packet suite: 1,353 passed
• invariant performance lane: 13 passed; allowance_read_p95_ms 592.888 ms recorded as invariants_only noise
• Pyright: 0 errors, 0 warnings
• release safety and dist checks passed
• wheel 373,870 bytes; sdist 819,975 bytes

Review

One comprehensive read-only reviewer completed. All three accepted findings were remediated: exact publication workload receipts, complete lifecycle-fold identity checks and mutation resistance, and preservation of the frozen preparation digest.

Scope

CK-07R1 lifecycle preparation only. PublicationWriter, recovery, schema, query, evidence service, and shared authority paths are unchanged.

[douglasmonsky]

CK-07R1 bounded blocker closeout: hosted CI run 30685780055 failed only in Python 3.14 job 91331138768. The exact failure is the pre-declared timing-sensitive Candidate A mandatory workload ordinary.2000_call_tail; the job otherwise completed 1,387 passing tests. Python 3.10 isolation and Focused Evidence Console passed. Per the packet stop condition, this PR remains open and unmerged; no gate, lifecycle invariant, publication invariant, or frozen budget was weakened, and no successor was created.