fix(preprocess): skip article fetches for incomplete discovery

[pengpengyi92]

Summary

• Stop PR Newswire collection before article-body fetching when listing discovery is incomplete.
• Preserve discovery failures and the observed row count while returning no partial documents, so completeness-preserving retries do not download the same parent-window articles first.
• Document the incomplete-window contract and add a focused regression test proving the article collector is not invoked.

This PR deliberately addresses only the redundant-fetch defect. The independent proposal to expose public article-fetch concurrency remains out of scope so that any configuration API can be reviewed and benchmarked separately.

Related Issue

Part of #144

Verification

• bash scripts/verify.sh in the project's Linux test environment: 486 passed, 86.46% coverage, import contracts and static checks passed.
• python -m pytest tests/preprocess/test_pr_newswire.py tests/flows/test_news.py tests/configs/test_news.py -q -o addopts=: 19 passed.
• python scripts/verify_news_e2e.py: PR Newswire discovery passed (332 observed rows across 4 pages) and ticker recovery passed (8/8 expected tickers); the independent RSS smoke check failed because the configured public PR Newswire RSS URL currently returns HTTP 404.
• A live 14-day collection probe reached the 2,000-row discovery cap and returned complete=False, documents=0, and the discovery failure evidence without collecting partial article bodies.

Checklist

• The title uses English Conventional Commit format: type(scope): summary.
• The related issue or design discussion is linked when applicable.
• bash scripts/verify.sh passes.
• Every applicable live-network component smoke test passes, or this PR states why none applies.
• Public behavior has focused tests, an example, and documentation where applicable.
• The PR is complete, small, and contains no unrelated changes.