fix(i18n): verify R2 publish before full translation batches

[hxy91819]

Summary

• replace async docs deploy dispatches in translation finalizers with a helper that dispatches r2-pages.yml, resolves the run, waits for completion, and fails on non-success
• make the full-translation canary commit/publish a one-page channels/line.md probe and live-smoke /zh-CN/channels/line for <h1>LINE</h1> before later full batches can run
• add workflow_dispatch canary_only=true so we can validate the canary publish path from this branch without starting the full translation batches
• guard canary artifacts/commits to the sampled page plus locale translation memory so stale pruned deletes cannot leak into the probe

Validation

• python .github/scripts/i18n/tests/test_i18n_scripts.py
• python .github/scripts/i18n/workflow_shell_check.py --workflows-dir .github/workflows --check-bash
• python .github/scripts/i18n/budget_check.py --workflow .github/workflows/translate-all.yml
• git diff --cached --check
• autoreview --mode local --engine codex --prompt-file I18N_CANARY_DEPLOY_FINDINGS_2026-06-27.md reported no accepted/actionable findings

Canary plan

After this PR is open, run Translate Full from this branch with target_locale=zh-cn and canary_only=true only. The expected proof is that the canary waits for r2-pages.yml and the live page https://docs.openclaw.ai/zh-CN/channels/line reports <h1>LINE</h1>.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.