Fix PR CI runner allocation on dzianisv/codex fork

[OpenCodeEngineer]

Summary

PR CI on dzianisv/codex is currently blocked by runner allocation and billing configuration before most jobs even reach repository code.

Observed on PR #23 (a725e6b1c) on 2026-03-16:

• rust-ci run 23167224186
• Bazel (experimental) run 23167224208

Failures observed

Missing self-hosted runner group

rust-ci job Tests — ubuntu-24.04 - x86_64-unknown-linux-gnu failed immediately with:

• Required runner group 'codex-runners' not found

The same workflow fanout shows multiple jobs completing with no steps executed and runner_id = 0, which is consistent with runner allocation failure rather than a code/build failure.

Billing / larger-runner allocation failure

rust-ci job Lint/Build — macos-15-xlarge - aarch64-apple-darwin failed immediately with:

• The job was not started because recent account payments have failed or your spending limit needs to be increased. Please check the 'Billing & plans' section in your settings

Bazel (experimental) macOS jobs also failed before executing steps, consistent with unavailable larger-runner capacity on the fork.

Why this matters

This blocks validation for unrelated PRs. On PR #23 the Rust/model-switching fix itself was validated locally, but the remote matrix could not actually run against the code because the fork's Actions infrastructure never started those jobs.

Acceptance criteria

• PR workflows on dzianisv/codex can allocate the configured runners successfully
• Any required self-hosted runner group references are valid for the fork, or workflows fall back to available runners on forks
• Larger-runner jobs do not fail at queue time due to account billing/spending configuration
• A normal PR can execute the rust-ci and Bazel (experimental) matrices past job startup