fix(crew-state): gate terminal PR read on real gh state

[lukeforshort]

Intent

Fix false-terminal read in bin/fm-crew-state.sh: gate the 'passed: merged/closed' report on real gh PR state instead of inferring it from a no-mistakes outcome label that also covers checks-passed-awaiting-merge, so supervision never ends on unlanded work.

What Changed

• Gated fm-crew-state.sh's outcome=passed terminal reading on a direct gh pr view check via a new pr_landed_state helper: merged/closed reports done, an open PR reports working (unlanded, awaiting merge), and an unresolvable/gh-unavailable state degrades to done (merge state unverified), so supervision is never ended on unlanded work by a false terminal label. Refactored the bounded no-mistakes runner into a generic run_bounded reused for the gh call.
• Hardened fm-watch.sh singleton-lock release by adding explicit TERM/INT/HUP traps alongside the existing EXIT trap, so a checkpoint killed by timeout's SIGTERM deterministically frees the lock instead of leaving a stale "already running" holder.
• Hardened fm-crew-state, fm-session-start, and fm-dispatch-select tests against environment assumptions and synced AGENTS.md, CONTRIBUTING.md, and the docs/ references for the gh-gated terminal read.

Risk Assessment

✅ Low: A tightly-scoped, well-tested fix that adds a real gh state check before a terminal supervision read, is safe by construction (open→working, unverifiable→done-without-merged-claim), backstopped by teardown's own landed-work verification, and introduces no correctness regressions.

Testing

Completed 1 recorded test check.

• Outcome: ⚠️ 1 error across 4 runs (1h4m46s)

Pipeline

Updates from git push no-mistakes

✅ **intent** - passed

✅ No issues found.

✅ **Rebase** - passed

✅ No issues found.

⚠️ **Review** - 2 infos

• ℹ️ bin/fm-crew-state.sh:508 - During the passed-but-still-open awaiting-merge window, every fm-crew-state.sh read for that crew now issues a bounded gh pr view API call (previously the outcome=passed path was purely log/pane-local). The watcher already runs a dedicated merge-detection poll (fm-pr-check's check.sh), so this is somewhat redundant network work on the supervision read path. It is bounded by NM_TIMEOUT (10s) and only fires on outcome=passed, so it is acceptable, but worth being aware of if crew-state read frequency grows.
• ℹ️ bin/fm-crew-state.sh:511 - When gh is unavailable/unauthenticated the outcome=passed path degrades to done (merge state unverified), i.e. the pre-fix behavior of reading a possibly-open PR as terminal. This is a deliberate, documented tradeoff and is backstopped by fm-teardown.sh's independent landed-work check, so unlanded work still cannot be torn down; noting it only so the protection's dependence on gh availability is explicit.

⚠️ **Test** - 1 error

• 🚨 tests failed with exit code 1
• command -v tmux >/dev/null || { echo "tmux is required for e2e tests" >&2; exit 1; }; tmux -V; rc=0; for t in tests/*.test.sh; do echo "== $t =="; bash "$t" || rc=1; done; exit "$rc"

🔧 Fix: harden session-start and dispatch-select tests against env assumptions
1 error still open:

• 🚨 tests failed with exit code 1
• command -v tmux >/dev/null || { echo "tmux is required for e2e tests" >&2; exit 1; }; tmux -V; rc=0; for t in tests/*.test.sh; do echo "== $t =="; bash "$t" || rc=1; done; exit "$rc"

🔧 Fix: crew-state tests pass; remaining failures are herdr-env and flaky watcher-lock
1 error still open:

• 🚨 tests failed with exit code 1
• command -v tmux >/dev/null || { echo "tmux is required for e2e tests" >&2; exit 1; }; tmux -V; rc=0; for t in tests/*.test.sh; do echo "== $t =="; bash "$t" || rc=1; done; exit "$rc"

🔧 Fix: release watcher singleton lock on terminating signals, not just EXIT
1 error still open:

• 🚨 tests failed with exit code 1
• command -v tmux >/dev/null || { echo "tmux is required for e2e tests" >&2; exit 1; }; tmux -V; rc=0; for t in tests/*.test.sh; do echo "== $t =="; bash "$t" || rc=1; done; exit "$rc"

✅ **Document** - passed

✅ No issues found.

✅ **Lint** - passed

✅ No issues found.

✅ **Push** - passed

✅ No issues found.

[kunchenguid]

Thanks for the PR! This branch currently has a merge conflict with the base branch.

When you get a chance, please rebase onto (or merge) the latest base branch, resolve the conflict, and push. After that, checks will re-run and the PR will get looked at again.

_{Noted for firstmate#404 at a93bc007.}