feat: improving performance and robustness

[PeterStaar-IBM]

Summary

This PR improves parsing throughput and memory behavior by reducing expensive font-resource JSON materialization, reusing already sanitized cells for
line-cell generation when the decode settings are equivalent, and avoiding redundant cell processing work. It also adds finer-grained timing keys for page
decoding, font decoding, sanitization, and word/line cell creation so performance analysis can identify the remaining hot spots more directly.

The build now forces Blend2D release assertion behavior for non-Debug artifacts and adds wheel checks to prevent shipping Blend2D debug assertion code.
The performance tooling was updated with better timing breakdowns, resumable comparison runs, PDF selection, parse/render content options, and refreshed
Apple M3 Max benchmark reports.

Changes

• Read font metadata directly from QPDFObjectHandle instead of eagerly converting font dictionaries to JSON.
• Add reusable QPDF object access helpers.
• Reuse sanitized cells for line-cell output when the requested line-cell config matches the sanitized representation.
• Split parser timings into more detailed nested buckets.
• Avoid extra work when extracting one-page PDFs from one-page documents.
• Add Blend2D release-mode compile/link settings and CI wheel assertion checks.
• Improve benchmark scripts with config controls, reusable comparison CSV data, PDF filtering, and clearer timing breakdowns.
• Add/update performance benchmark documentation and plots.

[github-actions]

✅ DCO Check Passed

Thanks @PeterStaar-IBM, all your commits are properly signed off. 🎉

[mergify]

Merge Protections

🟢 Merge protection satisfied — ready to merge.

Show 1 satisfied protection

🟢 Enforce conventional commit

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?:

[dolfim-ibm]

lgtm