docs: Document OpenAI review validation matrix

[idy]

Background

• Follow up to:
  • ci: Add issue-led PR readiness review gate #10

PR #11 introduced three fixed OpenAI review Checks and content-addressed PR, per-Issue, and code evidence. The implementation documentation explains rollout steps, but it does not give maintainers a compact event-to-stage validation matrix for verifying the default-branch caller after installation.

This follow-up is also the first end-to-end consumer test after the new caller and reusable reviewer reach main.

Goal

Document a live validation matrix for the Issue-led OpenAI review workflow in README.md.

The matrix must identify the expected trigger, affected evidence stage, cache mode, and fixed Check for:

• opening a ready PR;
• editing only the PR body;
• editing one linked Issue;
• pushing a descendant commit;
• rerunning an unchanged head.

This Issue owns documentation and live workflow validation only. It does not change reviewer behavior, configure a ruleset, merge the test PR, or roll the workflow out to another repository.

Code Changes Tree

README.md    # add the default-branch trigger and stage-cache validation matrix

Design

Add a concise Markdown table near the existing rollout guidance.

Each row must state:

• the GitHub event or manual action;
• which of PR, Issue, and Code evidence changes;
• the expected full, incremental, or reused stage behavior;
• which of OpenAI PR Review, OpenAI Issue Review, and OpenAI Code Review is expected to be recalculated.

The surrounding text must clarify that:

• all three fixed Checks are attached to the exact current PR head;
• Issue edits revalidate plan conformance without resending an unchanged complete code diff;
• unchanged reruns consume zero model tokens;
• a workflow-source or policy change intentionally invalidates compatible session evidence and starts the affected stages safely.

The PR body must natively close this Issue and list the validation run links. Live testing may update only this Issue's acceptance checkboxes and the PR description; it must not change reviewer code.

Test And Acceptance Criteria

• README.md contains the five-row trigger/cache validation matrix.
• Markdown remains readable and git diff --check passes.
• Opening the ready PR automatically creates OpenAI PR Review, OpenAI Issue Review, and OpenAI Code Review on the exact head.
  • Evidence: https://github.com/GizClaw/github-workflows/actions/runs/30509121677
• Editing only the PR body reruns OpenAI PR Review incrementally while unchanged Issue and Code evidence is reused.
  • Evidence: https://github.com/GizClaw/github-workflows/actions/runs/30509756489
• Editing this Issue reruns only its Issue stage incrementally and revalidates plan conformance without resending the complete code diff.
  • Evidence: https://github.com/GizClaw/github-workflows/actions/runs/30509613641
• Pushing a descendant documentation commit reuses PR/Issue evidence and reviews only the new code diff.
  • Evidence: https://github.com/GizClaw/github-workflows/actions/runs/30509999655
• Rerunning an unchanged head reports zero model tokens for PR, Issue, and Code evidence.
  • Evidence: https://github.com/GizClaw/github-workflows/actions/runs/30510100179
• The native ## 🤖 OpenAI PR review report shows all three verdicts and per-stage token/cache usage.
• The test PR remains open for review; this task does not merge it.