cli: Announce recovery secret switch

[BenWestgate]

What

When a complete valid secret is supplied after one or more shares were accepted, keep the existing deliberate behavior of ending share recovery and using that secret, but state that choice explicitly before returning it.

Why

The adversarial review correctly reproduced a silent mode switch. The switch itself is intentional; the defect is that already-accepted shares appeared to vanish without explanation.

Validation

• focused interactive recovery regression: 3 passed;
• full suite: 876 passed;
• full suite under python -O: 876 passed;
• Ruff check/format: clean;
• strict mypy across 21 source files: clean;
• git diff --check: clean.

Refs #38.

AI assistance: this focused follow-up is agent-authored. A responsible human should review and rewrite/squash it under the repository authorship policy before integration.

Current stack: head cebecfc is the same reviewed one-commit patch replayed directly on #81 (aa2d333); its stable patch-id is unchanged. The resulting library/CLI tip is 5,190 logical source lines. Human review order is #42 → #57 → #105 → #80 → #81 → #95.

[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]

Agent release-gate review at exact head 23dae08:

ACK on behavior. The adversarial-review finding was the silent mid-recovery mode switch: after compatible shares are already accepted, supplying a complete secret intentionally supersedes them. This patch preserves that policy and emits the notice only when accepted is non-empty, immediately before returning the complete secret.

Verification:

• exact diff is limited to two production lines plus one regression assertion;
• the existing compatible-share/secret regression passes 3/3 normally and 3/3 under python -O;
• git diff --check is clean;
• exact-head GitHub matrix is green and there are no review threads.

No code blocker found. Integration blocker remains authorship policy only: this is a Codex-authored commit and should be human-reviewed/re-written or squashed under the responsible human author before merge, as the PR body already records.

[BenWestgate]

AI-assisted review performed at the maintainer's request and disclosed per docs/developer/AI_POLICY.md.

No correctness findings. The change is narrowly scoped: it preserves the existing recovery behavior when a complete valid secret is entered after shares, but makes the mode switch explicit before returning the secret. The notice is conditioned on accepted, so it does not add noise when the secret is the first input. Focused recovery/secret tests pass locally (17 passed), and the exact-head GitHub matrix is green. Ready for human review/rewrite-squash under the repository authorship policy.

[BenWestgate]

Exact-head release-gate re-review: ACK 4ea72bb. This remains the same reviewed one-line behavior change by stable patch-id: when a complete valid secret is supplied after accepted shares, the CLI explicitly announces that it is switching to that secret; first-input secret recovery stays quiet. The exact integrated tip passes 935 tests normally and optimized, including the focused recovery-switch regression, and remains 5,193 <5200. No code blocker found. The agent-authored commit requires responsible-human rewrite/squash before integration.