fix: filter truncated candidates from batched best-of-n sampling

[liuzhengyang699]

The sequential fallback skips length-truncated completions, but the batched n path accepts them. A judge can then rank an incomplete answer above the complete candidates. Whitespace-only responses can also reach the judge.

Apply the same validity rule to both generation paths: keep nonblank content only when the finish reason is not length. An entirely invalid batch uses the existing sequential fallback.

Validation: 9 tests passed across tests/test_bon_candidate_filtering.py and tests/test_approaches.py. Three new cases fail before the fix; they verify the selected answer, which candidates reach the judge, fallback behavior, and token accounting for successful responses. Ruff checks on the new test and git diff --check passed.

[CLAassistant]

All committers have signed the CLA.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}