ci: run OCR models in one review job

[keIIy-kim]

Summary

• Rename the OCR workflow/check to the shorter OCR / review
• Install OpenCodeReview once per PR review job
• Run configured OCR models sequentially and post model-tagged comments

Validation

• ruby -e 'require "yaml"; YAML.load_file(".github/workflows/ocr-review.yml")'
• node --check /tmp/ocr-post-script.js
• actionlint .github/workflows/ocr-review.yml

[github-actions]

🔍 OpenCodeReview found 1 issue(s) in this PR.

• ✅ 1 posted as inline comment(s)
• 📝 0 posted as summary

[github-actions]

This step-level env declaration is redundant because job-level env variables are automatically inherited by all steps, including actions/github-script. More importantly, if the job-level OCR_LLM_MODELS variable is ever renamed or removed, ${{ env.OCR_LLM_MODELS }} will silently evaluate to an empty string, causing the script to process zero models with no visible error.

Consider removing this redundant env block entirely (the job-level env is sufficient), or referencing the variable directly via ${{ vars.OCR_LLM_MODELS || 'glm-5.2-fp8 kimi-k2.7-code' }} to preserve the default fallback and make the dependency explicit.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}