Prevent shell-echo races in automation docs

[tony]

Summary

Replace echo-prone completion polling with tmux wait-for channels when automation controls the command and does not need a timeout.

For bounded waits, retries, queues, and state machines, match echo-safe complete output lines against monotonic deadlines. Stop retry and queue sequencing when a timed-out foreground process may still own the pane.

Update every affected executable example and the page guidance so doctests prove command output rather than shell input echo.

Verification

• uv run ruff format --check .
• uv run ruff check .
• uv run mypy .
• just build-docs
• uv run py.test --reruns 0 -q docs/topics/automation_patterns.md (16 passed)
• uv run py.test --reruns 0 -q (1342 passed, 1 skipped)

Closes #706

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 52.46%. Comparing base (d8b7127) to head (2aff5ac).
⚠️ Report is 4 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #709      +/-   ##
==========================================
+ Coverage   52.44%   52.46%   +0.02%     
==========================================
  Files          26       26              
  Lines        3726     3726              
  Branches      747      747              
==========================================
+ Hits         1954     1955       +1     
  Misses       1468     1468              
+ Partials      304      303       -1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[tony]

Pushed a follow-up: check_for_errors was still carrying the bug this PR exists to remove.

The rest of the page now waits on wait-for channels and matches whole output lines, but that one helper was left doing pattern in '\n'.join(pane.capture_pane()). send_keys types the command, the shell echoes it back into the same buffer, so it reported a crash for any healthy command whose own arguments mention an error. Reproduced against live tmux before and after:

pane contents: ["true --log-level=ERROR; printf 'out%s%s\n' : ok; tmux wait-for -S ch", 'out:ok']

  before (scans whole pane):     'ERROR'   <- healthy command reported as a crash
  after  (scans tagged output):  None

The helper now takes the command's own output lines rather than a pane, so it cannot be handed the echo in the first place.

One detail worth flagging, because it caught me: the tag has to survive wrapping. My first attempt sent printf 'out:%s\n' ok and filtered on line.startswith('out:'). That passed — but only by luck. A long command wraps, and the wrapped continuation of the echo is its own line, so a tag that merely sits mid-line in what you typed can still end up starting one. The fix is to let printf assemble the tag from a format and a separate argument (printf 'out%s%s\n' : ok), which keeps the literal out of the echo entirely, wrapped or not. That is the same trick the page already uses for printf 'Success%s\n' '!', just applied to the prefix instead of the payload.

Also added a CHANGES entry under ### Documentation — the racy recipes shipped in 0.61.0, so they pass the Published-Release Test.

CI green on all 8 tmux versions.

Scope note: Closes #706 is now earned for the page #706 names, but the identical pattern is still live on two other pages — docs/topics/pane_interaction.md:500 and :519-521, and docs/topics/workspace_setup.md:399. Filed separately so it isn't lost when this merges.