[CI] Fix baseline QSA typing and pre-commit failures

[Leonccaa]

Baseline pre-commit failures block otherwise unrelated changes, including #617. This PR extracts the baseline CI fixes from that cache-retention PR so they can be reviewed independently.

• Correct _validate_qsa_e4m3_scale_load's return annotation to set[str], matching its implementation and callers; no scale-loading behavior changes.
• Exclude the existing AGPL-3.0-only scale-gate test from the Apache-specific SPDX header fixer. Preserve its existing license; all other checks and functional tests still apply.
• Remove the duplicate blank line in the current README and apply Ruff formatting adjacent to the corrected annotation.

Validation: VLLM_TARGET_DEVICE=cpu ... python -m pytest -p review_cpu_plugin tests/compile/test_qwen4exp_qsa_e4m3_gate.py -q: 5 passed. The local CPU plugin applies the existing skip_global_cleanup marker to avoid accelerator-only teardown; no functional tests are skipped. These exact three file contents previously passed full pre-commit (including Python 3.10–3.13 mypy) in #617 at fdc0ffc896, GitHub run 34767999595. Independent-branch CI will run on this PR; the previous run is historical evidence, not a claim about this new head.

No cache-retention changes or GPU runtime changes. Checked open PRs for overlapping baseline CI fixes; this extracts the changes from #617 rather than maintaining duplicate fixes. AI assistance: OpenAI Codex.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-13T16:30:14.465657Z	63fac24	PR opened

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.