Add --ci flag for headless review in CI/CD pipelines

[fuleinist]

Problem

The git lrc review flow is tightly coupled to a browser UI — lrc always opens a web view. Teams using AI-generated code in automated PR workflows (GitHub Actions, GitLab CI, etc.) have no way to integrate git-lrc checks as a CI gate. The review cannot run in a headless context.

Solution

Add a --ci flag to git lrc review that:

• Runs the review entirely headless — no browser opens, no UI
• Streams JSON output to stdout with structured results:

  {
    "issues": [
      { "file": "src/utils/auth.go", "line": 42, "severity": "high", "message": "Hardcoded API key detected" }
    ],
    "summary": { "total": 3, "high": 1, "medium": 1, "low": 1 },
    "coverage": 78,
    "iterations": 2
  }

• Exits with code 0 (clean) or 1 (issues found) for CI script gating
• --ci --fail-on=high only exits non-zero on high-severity issues (medium/low are warnings only)
• --ci --connector=openai allows specifying which BYOK connector to use in CI

Example GitHub Actions usage:

- name: Run git-lrc review
  run: |
    git lrc review --ci --fail-on=high > review.json || true
    echo "result=$(cat review.json)" >> $GITHUB_STEP_SUMMARY

Use case

Hooking git lrc review --ci into GitHub Actions or GitLab CI to enforce mandatory review on commits created by AI coding agents in automated PR workflows. Lets teams audit AI-generated diffs in CI without a human in the loop for every commit.

Scope

• New --ci flag handled in the existing review command handler
• Output format: JSON to stdout (no UI/web server)
• Exit codes: 0 = clean, 1 = issues found, 2 = error
• No new API calls beyond normal review flow
• Target: small and focused — purely a new output mode for existing logic