fix: enforce BFS max_pages across each level

[nightcityblade]

Summary

Fixes #2338

Count URLs already queued for the next BFS level against the remaining max_pages budget so sibling parent pages cannot oversubscribe it.

List of files changed and why

• crawl4ai/deep_crawling/bfs_strategy.py - Subtract the current next-level queue length when calculating remaining capacity.
• tests/deep_crawling/test_deep_crawl_resume.py - Add a regression test with two parents sharing the final two-page budget.

How Has This Been Tested?

• uv run --with pytest --with pytest-asyncio pytest tests/deep_crawling/test_deep_crawl_resume.py -q (33 passed)
• ruff check --select E9,F63,F7,F82 crawl4ai/deep_crawling/bfs_strategy.py tests/deep_crawling/test_deep_crawl_resume.py (passed)

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• My code is commented where needed — N/A; the capacity calculation and regression test are self-explanatory
• I have made corresponding changes to the documentation — N/A; this corrects existing max_pages behavior without changing the public API
• I have added/updated unit tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

[ntohidi]

Thanks for picking this up so quickly, @nightcityblade, and for adding a test!

#2339 fixed the same issue and is now merged. It includes your len(next_level) change and also caps each level before it is fetched, which the batch mode needed to stop at exactly max_pages. So I'm closing this one as a duplicate.

Thanks again for your effort, and we'd be happy to see more PRs from you.