fix(i18n): carry translation memory in full shards

[hxy91819]

Summary

• keep locale translation memory in shard 0 artifacts for sharded full translation runs
• add a regression test for shard 0 vs non-zero shard packaging
• document the temporary recovery checklist for R2 publish, targeted hi/ru recovery, and finalizer/commit failures

Verification

• python .github/scripts/i18n/tests/test_i18n_scripts.py
• actionlint -config-file .github/actionlint.yaml
• 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 --check upstream/main...HEAD
• .agents/skills/autoreview/scripts/autoreview --mode branch --base upstream/main

[chatgpt-codex-connector]

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