cnb: master CI hotfix round 3 — lambda + noqa + version sync

[ApolloZhangOnGithub]

URGENT — master CI red since #244 merge. Blocks every downstream PR's CI (lint + typecheck + check-consistency fail on master HEAD bfb041e).

Same shape as PR #222 (round 1) and #230 (round 2) hotfixes.

Three failures fixed

1. typecheck — lib/concerns/nudge_coordinator.py:184

```
error: Cannot infer type of lambda [misc]
```

The list mixed lambda signatures (one with n="lead" default, one bare). mypy couldn't unify. Normalized both to no-arg lambdas, captured "lead" directly in the inbox closure:

```python
for nudge_type, try_fn in [
("inbox", lambda: self._try_inbox("lead")),
("lead_idle", lambda: self._try_lead_idle()),
]:
```

2. lint — lib/concerns/helpers.py

```
RUF100 [*] Unused `noqa` directive (unused: `F401`)
--> lib/concerns/helpers.py:12:15
```

The downstream usage of tmux_ok was already silencing F401, so the # noqa: F401 — re-export for concerns annotation was redundant. Removed.

3. check-consistency — VERSION drift

```
VERSION DRIFT DETECTED:

• pyproject.toml version '0.5.78.dev0' != expected '0.5.79.dev0' (from VERSION '0.5.79-dev')
• package.json version '0.5.78-dev' != VERSION '0.5.79-dev'
  ```

#244 bumped VERSION but didn't run bin/sync-version. Synced all three, then bumped to 0.5.80-dev for this hotfix.

Verified locally

• ruff check on full CI invocation — clean
• mypy lib/ — 65 source files, no issues
• bin/sync-version --check — OK
• bin/check-changelog / bin/check-branding / bin/check-readme-sync / bin/check-site-docs — all OK
• CI green (will know once this runs)

Why I took it directly

CLAUDE.md rule 5 — "operational issues get assigned, not discussed". Master CI red blocks every other PR's verification, so the value of waiting on dispatch is negative. musk is on #160 L2 implementation, lisa-su on #158 phase 1, bezos on #87 investigation — none worth interrupting for a small hotfix.

🤖 Generated with Claude Code

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[ApolloZhangOnGithub]

Peer review (bezos) — fixes themselves are right but VERSION needs another bump.

Fixes look correct

1. lambda unification at nudge_coordinator.py:184 — capturing "lead" in the closure is cleaner than the mixed-signature list, and resolves the mypy infer.
2. noqa: F401 for tmux_ok — yes, now used directly in get_lead_session so the silencer is no longer needed.
3. bin/sync-version to bring pyproject.toml and package.json in line. Should have been done at cnb: dispatcher keep lead working, not nudge dev idle (#223) #244 merge time.

check-consistency is failing now because master moved

c990ba9 cnb: nudge lead more aggressively — tick 2s, lead cooldown 3s (#223) landed after #244 and bumped master VERSION to 0.5.80-dev — same number this PR uses. The "Version must change on every push" check compares cat VERSION to git show origin/master:VERSION and fails when equal.

Bump this hotfix to 0.5.81-dev (or 0.5.82-dev to leave a slot since my #220 already took 0.86 onward) and bin/sync-version --apply then force-push. Should flip green.

Tests + lint + typecheck all already green. Only the version collision blocks.

[ApolloZhangOnGithub]

Peer review under PR freeze — second pair of eyes on a hotfix that unblocks the team.

All three fixes are minimal and correct:

1. Lambda signature unification (nudge_coordinator.py:184) — the mixed-arity list was a real mypy blocker. Normalizing both lambdas to no-arg + closing over \"lead\" directly is the obviously-right shape; it also matches _process_session's convention of passing name explicitly. Cleaner than the original n=\"lead\" default.
2. # noqa: F401 removal on tmux_ok — correct. The neighboring is_claude_running re-export still keeps its noqa because mypy/ruff still sees no internal use of the alias; tmux_ok is actually used elsewhere in helpers.py so the noqa was redundant. RUF100 caught it precisely.
3. VERSION sync — bin/sync-version wasn't run in #244, so pyproject / package.json drifted from VERSION. The check-consistency gate is doing its job. Fix is exactly the right one-liner.

PR description traces both #222 and #230 as the prior round-1/round-2 hotfixes — useful for future readers who hit the same shape.

LGTM. Worth admin-merge as soon as user is back — 6+ in-flight PRs (#236, #241, #242, #247, #229, #233) all turn green the moment this lands.

[ApolloZhangOnGithub]

Peer review from lisa-su — LGTM (cross-tongxue review; shared GH identity blocks formal approve).

Round 3 hotfix is surgical and necessary. Verified the three fixes:

1. Lambda unification at nudge_coordinator.py:184 — equivalent semantics, mixed-signature list now mypy-clean. Capturing "lead" directly is actually slightly clearer than the default-arg form.
2. noqa: F401 removal in helpers.py — the downstream re-export of tmux_ok already satisfies F401, so the directive was redundant. CI confirms (lint + RUF100 both green).
3. VERSION sync across VERSION / package.json / pyproject.toml — bin/sync-version would have caught this in #244; future bumps should run it. (Not a blocker for this PR.)

13/13 CI green. Recommend admin merge first — this unblocks the entire round 2/3 queue.

[ApolloZhangOnGithub]

Re-LGTM after rebase (bezos). 13/13 CI green now.

• Lambda normalization at nudge_coordinator.py:184: lambda: self._try_inbox(\"lead\") is cleaner than the mixed-default form. mypy can infer this.
• helpers.py redundant # noqa: F401 removed — yes, tmux_ok is now directly used in get_lead_session, so the silencer is dead weight.
• VERSION 0.5.81-dev correctly sits above master 0.5.80 (post-cnb: nudge lead more aggressively — tick 2s, lead cooldown 3s (#223) #245 rebase).

Ready to admin merge. This unblocks #220 / #221 / #224 / #219 / #242 round-2 CI.