Update guardrail scripts to use hasHumanOverride

[jdelfino]

Summary

Replace the per-guardrail listReviews + hasNonStaleApproval pattern with the unified hasHumanOverride helper from override.js. Update user-facing messages to mention both override options.

Files to modify

• .github/agent-workflow/scripts/guardrail-scope.js
• .github/agent-workflow/scripts/guardrail-api-surface.js
• .github/agent-workflow/scripts/guardrail-test-ratio.js
• .github/agent-workflow/scripts/guardrail-dependencies.js
• .github/workflows/guardrail-commits.yml (inline script)
• .github/agent-workflow/scripts/orchestrator-check.js (messaging only)

Implementation steps

1. In each of the 4 guardrail scripts (scope, api-surface, test-ratio, dependencies):

   • Replace require('./lib/approval.js') with require('./lib/override.js')
   • Replace the listReviews + hasNonStaleApproval block with const hasValidApproval = await hasHumanOverride({ github, owner, repo, prNumber, headSha })
   • Update override instruction messages: "approve the PR to override" → "approve the PR or comment LGTM to override"
   • Update "approved by reviewer" titles to "overridden by human" or similar

2. In guardrail-commits.yml inline script:

   • Replace require('./lib/approval.js') with require('./lib/override.js')
   • Replace review fetch + check with hasHumanOverride call
   • Update messaging

3. In orchestrator-check.js:

   • Update the message on line 215 to mention LGTM as alternative

Acceptance criteria

• No guardrail script imports approval.js directly (all use override.js)
• All 5 guardrails + commit workflow check both PR approval and LGTM comments
• User-facing messages mention both "approve" and "LGTM" options
• npm test passes

Dependencies

Blocked by #43