feat: --fail-on-warnings / failOnWarnings

[NormalGaussian]

yarn commands can be made to exit with a non-zero status code due to the presence of warnings logged during execution.

What's the problem this PR addresses?

Its a partial resolution to #7066

CI / testing workflows want to proactively avoid bugs and issues in the repo's they manage whilst being clear and easy to maintain themselves. Yarn emitting warnings is a clear indication of potential bugs and issues in the repo, but it is not clear and easy to detect such warnings. A simple grep is prone to false positives (e.g. a package name) and more complex solutions are difficult to understand, adequately maintain, and demonstrate as being free of bugs e.g.

yarn install --json | tee >(jq -c 'select( .type == "warning")' > warnings.jsonl) | jq -r '.indent + .data'
test "$(wc -l < warnings.jsonl)" -gt 0 && echo "Unfiltered warnings found in 'yarn install'" && exit 1

yarn install --fail-on-warnings is a lot simpler.

How did you fix it?

The CLI option --fail-on-warnings was added as an option to BaseCommand and its presence propagated to StreamReport.start({...}) as failOnWarnings. When StreamReport determines the exitCode it now takes into consideration the failOnWarnings setting and the presence of warnings.

Of note: the change was added to BaseCommand, which seemed the most appropriate as this is a generic behaviour, but it was added as a hidden option as it seemed to confusingly pollute yarn -h for most users, for what is generally a CI option. This has made it unclear as to how best to document.

Checklist

• I have read the Contributing Guide.
• I have set the packages that need to be released for my changes to be effective.
• I will check that all automated PR checks pass before the PR gets reviewed.

[NormalGaussian]

The failing checks appear to be the Error: EBADF: bad file descriptor, fstat that has been present since Node 25.7.x