ci: stop weekly arm64 mixed chaos from timing out

[fwerkor]

Summary

• run QEMU chaos shards on native architecture runners
• size mixed-chaos job timeouts for both sequential QEMU phases
• validate the production arm64 mixed-chaos limits on a native arm64 runner

Root cause

mixed_chaos.sh runs two sequential QEMU phases, each bounded by ARGOSFS_QEMU_TIMEOUT. The weekly arm64 shard allowed 90 minutes per phase but capped the whole job at 115 minutes. Healthy runs therefore reached phase 2 and were reported as cancelled when the outer job timeout expired. Running the arm64 guest under x86_64 TCG also made the mismatch repeatable.

Validation

• actionlint on both modified workflows
• YAML parsing with PyYAML
• git diff --check
• PR-only native arm64 mixed-chaos replay using a prior weekly arm64 image

Closes #132

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 62e9726ede

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[fwerkor]

The one-off native arm64 replay is running at https://github.com/fwerkor/ArgosFS/actions/runs/30689133228.

The previous native replay reached all phase-1 success markers and entered phase 2 before the old 115-minute outer timeout cancelled it. The final branch removes the one-off workflow so no hard-coded historical artifact run is retained in main; the production weekly workflow keeps the native runner selection and corrected two-phase timeout budgets.