Remove pre-review CLI dead code

[BenWestgate]

Remove the mechanical cleanup items recorded by the adversarial-review triage and reduce installed review surface without changing user-visible behavior:

• remove the redundant _create terminal guard;
• remove the permanently false correction ambiguity return field and unreachable consumer while preserving the mixed-case required-before-optional scheduling and capture accounting;
• update test doubles/mocks to the resulting three-value _correction_candidates() contract;
• remove the fake Bitcoin Core initialize(private=...) state that production does not expose;
• move correction helpers used only by tests/benchmarks from src/codex32/indel.py to tools/correction_reference.py;
• remove the tracked extra blank line in pyproject.toml.

Why this now precedes #57

The current Core-native reviewability-v1 base increased installed source enough that the focused #57 restore-authentication change reaches 5,228 production logical lines by itself, exceeding the maintainer-authorized <5200 gate. The already-reviewed mechanical cleanup is independent of that security behavior and reduces the #42 tip from 5,124 to 5,030 lines. Applying it first lets #57 remain focused and keeps every intermediate commit inside the existing budget without raising the cap or taking a risky refactor merely to save lines.

Current head 0867a1b is one focused Ben Westgate-authored cleanup commit stacked directly on #42 (4be79a7). Conflict resolution deliberately preserved #42's current wallet behavior; #57-specific record/fingerprint behavior is no longer bundled into this cleanup.

Validation

• installed production logical review lines: 5,030, below <5200;
• local exact-head suite: 898 passed normally;
• local exact-head optimized suite: 898 passed under python -O;
• all 232 CLI tests pass;
• strict mypy passes on all 21 source files;
• Ruff check and format check pass;
• correction constants verification passes;
• frozen differential correction verifier passes all 57 cases;
• git diff --check is clean.

Human review order is now #42 → #46 → refreshed #57. #80 and #81 follow #57. This ordering is mechanical: it exists to keep the authorized review-size invariant true at every step.

Intentional non-goals: no correction-policy change, wallet-identity change, helper/API publication, or fuzzing-policy change. Those are owned by #42/#45, #57/#81, #53, and #40/#47 respectively.

AI assistance was used for the user-authorized restack/conflict validation; the squash/restack does not substitute for responsible maintainer review.

[BenWestgate]

Agent handoff: before declaring the pre-review cleanup complete, read the ignored local checklist at docs/planning/v1-pre-review-cleanup.md. It records the three changes in this PR, validation commands, related gates #38–#41, and explicit non-goals. Please respond here if the local file is missing or any checklist item is not represented by commit 11e1a2d.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.

[BenWestgate]

AI-generated review (Claude), posted at the maintainer's request.

Concept ACK. Code looks right; not ACKing 7e7c649 until the merge commits are gone.

• Removed branches are unreachable: _create already requires a TTY on stdin/stdout, and _connected_core never returns None. ambiguous was always False.
• Rebase away 825b863/ccabfee.
• The PR body points to docs/planning/v1-pre-review-cleanup.md, which is local and ignored, so reviewers can't read it. Inline the non-goals or drop the pointer.
• Conflicts with #53 and #57.

[BenWestgate]

Review follow-up: the PR body no longer points reviewers at the ignored local planning file; the non-goals are now inline. The remaining NACK item is the branch history itself: rebase away the two base-refresh merge commits before human review. I’m leaving #46 behind #45 in the review order until that rewrite is done.

[BenWestgate]

Release-gate follow-up: the earlier history NACK is resolved. Current head 28b06c6 is one human-authored cleanup commit directly on #45 head eb7cfdd; the base-refresh merges and agent-authored commit are gone, and fresh Python package run 36333842809 completed successfully. The remaining dependency is sequencing, not a defect in this head: review/merge #45, then #57, then refresh this one cleanup commit once onto the integrated reviewability-v1 tip because it overlaps #57 in cli.py/tests/test_cli.py.

[BenWestgate]

Release-gate refresh note: final #45 is now c943e9a and #57 is eff541c0. I simulated reviewability-v1 → #42 → #45 → #57: all three merge cleanly. Reapplying this cleanup then conflicts only in src/codex32/cli.py and tests/test_cli.py, where the refresh must preserve #57’s restore-identity parameters/behavior while removing the dead cleanup paths. No final review requested until that one refresh is done after the prerequisites land.

[BenWestgate]

Final-refresh addendum: disposable replay of this single cleanup commit on the current combined #12/#13/#59/#7/#51/#33/#42/#45/#57 stack resolves only the expected cli.py/test_cli.py conflicts when replayed as a commit, but the resulting installed package is 5,098 logical review lines. That fails the documented <5000 v1 invariant. Tracked as #61.

Do not raise the budget in this PR. The post-prerequisite cleanup refresh must either recover at least 99 real review lines while preserving #42/#57 behavior or compose with a separate focused cleanup that does, then rerun the exact combined-tree size test.

[BenWestgate]

AI-generated review, posted at the maintainer's request.

Not ACKing 28b06c6 yet: this head is intentionally stale/conflicting. Refresh it after #42/#45/#57 land, then rerun validation.

[BenWestgate]

AI-assisted release-gate recheck at exact head 0793590. Code ACK. I reviewed the current one-commit delta after the correction helper split: it preserves _scheduled_candidates() and the required-before-optional/capture-ledger behavior, removes only the permanently-false ambiguity field and consumers, unreachable TTY/disconnected-Core branches, the fake-only private= test parameter, and reference-only indel helpers from installed source. Exact-head run 36801637475 is green and there are no unresolved review threads. No remaining code blocker in #46 itself; final integration still waits for human integration of #42, then the one-time #57 refresh. The commit is already Ben Westgate-authored.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.

[BenWestgate]

Release-gate ACK 0867a1b.

I re-reviewed the refreshed one-commit cleanup on current #42. The earlier nACK was mechanical/staleness only; current head removes the unreachable _create guards, permanently-false ambiguity field, drifted Core test-double state, and reference-only installed helpers without changing correction policy or wallet behavior. The branch is now one Ben Westgate-authored commit with no unresolved inline threads. Exact-head local validation passed 898 tests normally and under python -O, all 232 CLI tests, strict mypy, Ruff check/format, correction-constant verification, all 57 frozen differential cases, and git diff --check.

No code blocker remains from this review. The dependency-aware order is #42 → #46 → #57; placing this mechanical cleanup before #57 keeps the installed-source budget below the authorized <5200 cap at every intermediate tip.

[BenWestgate]

Agent exact-head review on 0867a1b:

• reviewed the full one-commit diff against correct: Interpret mixed-case damage #42;
• removed correction helpers are referenced only from tests/ and tools/, not installed production code;
• the correction ambiguity field was permanently false and its removal leaves the fail-closed incomplete/no-candidate behavior intact;
• the duplicate _create TTY guard is unreachable after the existing enclosing interactive-terminal check;
• affected local suite: 349 passed; git diff --check is clean; exact-head GitHub CI is green.

No code blocker found. One non-blocking commit-message nit remains: the commit body contains literal \\n\\n escape text (and a trailing quote) instead of paragraph breaks. Fix that when the human rewrites/signs the commit for integration; rewriting #46 solely for this nit now would unnecessarily invalidate the already-reviewed #57 stack.

[BenWestgate]

LGTM