feat: migrate to gobenchdata for continuous benchmarking with PR comments only

[Copilot]

This PR migrates the repository from a custom benchmark comparison workflow to use the gobenchdata action for continuous benchmarking, configured to publish results only as PR comments rather than to GitHub Pages.

Changes Made

🔄 Workflow Migration

• Replaced the existing benchmark.yml workflow with a gobenchdata-based implementation
• Added benchmark-publish.yml to establish baseline benchmark data on main branch pushes
• Configured gobenchdata to use checks mode for PR comments only (no GitHub Pages publishing during PR reviews)

🛠️ Comment Management

• Migrated from custom GitHub script to peter-evans actions:
  • Uses peter-evans/find-comment@v3 for finding existing benchmark comments
  • Uses peter-evans/create-or-update-comment@v4 for creating/updating PR comments
• Improved comment formatting with structured results and clear explanations

⚙️ Configuration

• Added gobenchdata-checks.yml with performance regression thresholds:
  • Performance (ns/op): 20% maximum regression allowed
  • Memory allocation (B/op): 25% maximum increase allowed
  • Allocation count (allocs/op): 15% maximum increase allowed
• Configured benchmark data retention (30 runs) and proper error handling

📚 Documentation

• Updated README.md to document the new continuous benchmarking system
• Replaced references to the old benchstat-based approach with gobenchdata information

💯 Conventional Commits

• Used conventional commit format throughout the implementation
• Configured gobenchdata to use conventional commits for benchmark data updates

Benefits

• More robust benchmark comparison with configurable performance thresholds
• Better error handling for scenarios with missing baseline data
• Cleaner PR comments with structured, easy-to-understand results
• Historical tracking of benchmark performance over time (via baseline publishing to gh-pages)
• Reduced maintenance overhead compared to the custom solution
• PR-only workflow that keeps benchmark comparisons in PR discussions without cluttering GitHub Pages

Testing

• ✅ Verified gobenchdata configuration works correctly with CLI testing
• ✅ Confirmed benchmark field names match actual JSON structure
• ✅ Tested that existing build and test workflows remain unaffected
• ✅ Added graceful handling for first-run scenarios without baseline data
• ✅ Removed GitHub Pages publishing from PR workflow per user request

The new workflow establishes baseline data when changes are merged to main, and subsequent PRs show performance comparisons in PR comments only without publishing to GitHub Pages during the review process.

---

💬 Share your feedback on Copilot coding agent for the chance to win a $200 gift card! Click here to start the survey.

[icedream]

@copilot Can you make it so using actions peter-evans/find-comment@v3 and peter-evans/create-or-update-comment@v4 we do not publish benchmark comparisons to GitHub Pages but instead to a comment on a PR again?

[Copilot]

@copilot Can you make it so using actions peter-evans/find-comment@v3 and peter-evans/create-or-update-comment@v4 we do not publish benchmark comparisons to GitHub Pages but instead to a comment on a ...

Done! I've updated the benchmark workflow to remove GitHub Pages publishing from the PR workflow. Now the benchmark.yml workflow only creates PR comments using peter-evans actions and doesn't publish to GitHub Pages.

The baseline benchmark data will still be published to the gh-pages branch when changes are merged to main (via benchmark-publish.yml), which provides the historical data needed for comparisons. Commit: 80153db