fix: reduce deduplication pair review effort to high

[kmbroai]

Summary

Reduce the reasoning effort for Sol pairwise duplicate comparisons from xhigh to high.

Changes

• Set gpt-5.6-sol pair reviews to high reasoning effort.
• Update the existing model/effort assertions, Docker smoke-test mock, and deduplication documentation.

Testing

• Focused deduplication tests: 17 passed.
• Docker mock request check: accepts Luna xhigh screening and Sol high pair reviews; fixture formatting passed.
• pnpm run types: passed.
• pnpm run format: passed.
• pnpm run test --seed 12345: 2,400 passed, 41 skipped, one unrelated authentication test failed because ambient API-key authentication overrides its expected ChatGPT login. That test passed with OPENAI_API_KEY and CODEX_API_KEY unset.
• pnpm run test with ambient API-key variables unset: 2,401 passed, 41 skipped, zero failures (seed 3693900373).

Risk and rollout

Pair-review decisions may differ with lower reasoning effort. Luna screening remains at xhigh; models, prompts, schemas, and CLI syntax are unchanged. Pair-review checkpoints include reasoning effort, so prior xhigh checkpoints will not be reused for new high requests.

Public disclosure review

• No customer, partner, prospect, or user identities, data, or identifying details are included.
• No credentials, personal data, private source, scan findings, or nonpublic links or tickets are included.
• I reviewed the branch name, title, description, commits, changes, comments, logs, screenshots, attachments, and links for public disclosure.

[chatgpt-codex-connector]

Codex Review Summary

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

Review	Status	Commit	Review trigger
🔒 Security Review	✅ Completed 2026-09-04T17:47:09.205974Z	81383d7	New commits

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.