feat(cloud): detect dead local daemon in cloud status and document launchd unit

[jlsevillano]

🔗 Linked Issue

Closes #279

---

🏷️ PR Type

• type:bug — Bug fix
• type:feature — New feature
• type:docs — Documentation only
• type:refactor — Code refactoring (no behavior change)
• type:chore — Maintenance, dependencies, tooling
• type:breaking-change — Breaking change

---

📝 Summary

• engram cloud status now probes the local engram serve daemon at 127.0.0.1:7437 (respects ENGRAM_PORT) with a 1s timeout and prints a Local daemon: running | not running | unreachable line so users can detect a silently dead autosync after brew upgrade engram or any other binary replacement.
• Adds a launchd template to DOCS.md "Running as a Service" so macOS users can supervise engram serve the same way Linux users do with systemd. With KeepAlive=true, autosync now survives brew upgrade automatically.
• Probe is informational: exit code is unchanged. Probe is suppressed when cloud is not configured to avoid noisy output for users who don't use cloud sync.

📂 Changes

File	Change
cmd/engram/cloud_daemon_probe.go	New: cloudDaemonProbe variable function (1s timeout GET /health), port resolution (ENGRAM_PORT → 7437), and printCloudStatusDaemonProbe writer with recovery hint when the daemon is down.
cmd/engram/cloud_daemon_probe_test.go	New: unit tests covering all four probe outcomes (running, refused, non-2xx, timeout) and the three printer states.
cmd/engram/cloud.go	cmdCloudStatus calls printCloudStatusDaemonProbe in each cloud-configured branch (token, token+insecure, no-token) before the existing sync diagnostic. No behavior change in the "not configured" branch.
cmd/engram/main_extra_test.go	stubRuntimeHooks now stubs cloudDaemonProbe so existing tests stay deterministic; new TestCmdCloudStatusEmitsLocalDaemonLine verifies the line is printed when configured (and suppressed when not).
DOCS.md	Renames Using systemd → Using systemd (Linux). Adds Using launchd (macOS) with full plist template (KeepAlive=true so brew upgrade does not break autosync), load/unload steps, and verification via engram cloud status. Updates the engram cloud status reference bullet to describe the new Local daemon: line.
docs/INSTALLATION.md	Homebrew section gains a tip pointing macOS users at the launchd template so autosync survives brew upgrade.

🧪 Test Plan

• Unit tests pass locally: go test ./... (passes with the standard CI environment; my local env had ENGRAM_CLOUD_SERVER set which leaks into pre-existing tests — repro by unset ENGRAM_CLOUD_SERVER ENGRAM_CLOUD_TOKEN ENGRAM_CLOUD_INSECURE_NO_AUTH ENGRAM_CLOUD_AUTOSYNC ENGRAM_PORT before running, same isolation as CI).
• E2E tests pass locally: go test -tags e2e ./internal/server/....
• Manually verified all three states with cloud config --server ...:
  • daemon down on port 7777 → Local daemon: not running on port 7777 + recovery hint mentioning engram serve and the launchd template
  • daemon up on port 7777 → Local daemon: running on port 7777
  • ENGRAM_PORT=9000 while daemon stays on 7777 → probe targets 9000 and reports not running on port 9000 (env override honored)

---

🤖 Automated Checks

These run automatically and all must pass before merge:

Check	What it verifies	Status
Check Issue Reference	PR body contains Closes #N / Fixes #N / Resolves #N	⏳
Check Issue Has status:approved	Linked issue has status:approved label	⏳
Check PR Has type:* Label	PR has exactly one type:* label	⏳
Unit Tests	go test ./... passes	⏳
E2E Tests	go test -tags e2e ./internal/server/... passes	⏳

---

✅ Contributor Checklist

• I linked an approved issue above (Closes #279)
• I added exactly one type:* label to this PR (type:feature)
• I ran unit tests locally: go test ./...
• I ran e2e tests locally: go test -tags e2e ./internal/server/...
• Docs updated (DOCS.md service section + cloud status reference + INSTALLATION.md tip)
• Commits follow conventional commits format
• No Co-Authored-By trailers in commits

---

💬 Notes for Reviewers

• The Homebrew formula post-install hook approach mentioned in the issue lives in homebrew-tap, so it is intentionally out of scope for this PR. The two in-repo mitigations (status probe + launchd template) close the gap on this side.
• The probe deliberately distinguishes not_running (TCP dial error to 127.0.0.1) from unreachable (timeout / non-2xx / unexpected error) so the recovery hint only fires when restarting engram serve is the right action.
• Probe timeout is exposed as a var daemonProbeTimeout so tests can shorten it; default in production stays at 1s.
• Plist uses literal <HOME> placeholders because launchd does not expand $HOME/~ inside plist values; the docs explicitly call this out.