ci: test-readme enhancements

[pskfyi]

See #47, #59.

[pskfyi]

After rebasing on top of main CI is accurately surfacing a unit test added by this PR which was passing within the PR, which was branched off of an old main, but which fails in the context of the updated main. This is the exact reason that we want to keep PRs in sync with the base branch. I disabled this setting before because the GitHub UI unfortunately defaults to a merge commit rather than rebasing for this. I think we should try enabling for a while and seeing if it causes more issues than it catches.

Fix for the failing test is incoming.

[AustinArey]

Updated branch via rebase. I see the force push above. Now evalCodeBlocks() test fails.

I did git fetch then git checkout test-readme-cross-platform then git pull and deno test -A md ALL PASS

I am confused why it fails in the PR actions if I have the up to date code locally.

[pskfyi]

There was a compiler error introduced by the final commit because #49 added the return type of Promise<void> to evalCodeBlocks (meaning it should return nothing or undefined) but the final commit here returned the results of evaluateAll from earlier in the function. Updating the return type resolved the issue 👍

[AustinArey]

Nice work thanks for the merge conflict fix.

[pskfyi]

Awesome 🎉

If you still have the failing code locally, you can run deno task test-readme and see that the evalCodeBlocks one fails. That's why it failed in CI in all 3 OS's.

Surprisingly, deno lint and deno fmt --check which we're running in CI don't surface the issue. I'm wondering if we should add a step to run deno check in Ubuntu, which does surface these sorts of issues. Right now if we had such an error in a file not imported in a readme example, or only imported in a readme example which uses no-eval, then the compiler error would not be surfaced in CI.