test(eval): exercise GitHub Actions reporter

[Abhijeet Prasad (AbhiPrasad)]

Summary

• add a self-contained eval that intentionally errors one case
• add a GitHub Actions workflow that runs the eval with --reporter=github-actions and verifies the workflow annotation
• capture case identity events from the JS SDK bridge so local/no-send-logs evals provide real case IDs instead of a degraded synthetic stream

User-facing test

The workflow prints a command like:

::error title=case 0 errored::intentional error from the GitHub Actions reporter example

GitHub renders that command as an annotation. The eval itself exits non-zero by design, while the workflow verifies both the annotation and expected exit status.

How to run locally

pnpm install --dir examples/eval-reporters --frozen-lockfile
cargo build --bin bt
cd examples/eval-reporters
../../target/debug/bt eval --no-send-logs \
  --reporter=github-actions \
  github-actions.eval.ts

Validation

• mise exec -- pre-commit run --all-files
• cargo test --test eval_golden
• local GitHub Actions reporter example

Stacking

This PR is stacked on #264 and should be reviewed after that reporter API PR.