docs: document Codex App backend protocol

[SSBrouhard]

What Changed

• Added docs/codex-app-backend.md describing the proposed Codex App visible-thread backend protocol.
• Linked the Codex App backend protocol from the architecture and configuration docs.
• Included pipeline-discovered Bash nounset fixes in fm-spawn.sh and fm-pr-merge.sh; the pipeline test stage initially caught this and reran clean after the fix.

Risk Assessment

✅ Low: this is a docs-only change that clearly marks codex-app as proposed and not selectable for new task spawns.

Testing

Completed 1 recorded validation pass.

• Outcome: 1 issue found → auto-fixed ✅ across 2 runs (25m52s)

Pipeline

Updates from git push no-mistakes

• ⏭️ intent - skipped
• ✅ Rebase - passed
• ✅ Review - passed
• 🔧 Test - 1 issue found → auto-fixed ✅
• ✅ Document - passed
• ✅ Lint - passed
• ✅ Push - passed

[kunchenguid]

Thanks for the PR! It looks like this branch has a merge conflict with the base branch right now. When you get a chance, could you rebase onto (or merge in) the latest base branch, resolve the conflict, and push? Once GitHub shows the PR as mergeable again, it'll be picked back up for review.

_{Noted for firstmate#211 at d410cf88.}

[kunchenguid]

Reviewed together with #212 - full reasoning in my comment there: #212 (comment).

Short version: the forward-looking protocol material in this doc - the host-tool catalog (create_thread / send_message_to_thread / archive), the identity-alignment discussion, the smoke expectations - is actually the right foundation, and it's the direction I want: a codex-app backend that meets the same create/send/observe/kill contract as tmux/herdr/zellij, not a passive ledger beside them.

Two mechanical problems with landing this as-is: it now drifts from what #212 actually implements, and the two PRs ship different files at the same docs/codex-app-backend.md path, so they conflict with each other in both merge orders. Please reconcile to one doc (this one's protocol/future material folded in as the contract an implementation must meet), and see the three concrete asks on #212.

[SSBrouhard]

Superseded by #222, which reconciles the Codex App backend contract into one docs path, keeps the useful host-tool protocol material, records the live Codex Desktop smoke, and answers the status-file return-channel question.