docs: add real world code review evaluation - opencode

[christso]

The code reviewer should have picked up on the opencode regression in anomalyco/opencode#7379.

Add an eval (which will later be part of an eval suite) for code reviewers across different models and agent skills.

We want to measure false positives, missed issues, review depth.

[christso]

Thanks, this older benchmark/example research slice is now folded into the public benchmark examples and best-practices roadmap in #1076 and #1137. Closing this issue as consolidated so public discussion stays on the broader benchmark documentation threads.