Add comprehensive tests for all new PDF features

Add core agentic PDF features: metadata, outline, search, links, images, forms

fix: resolve 8 real-world PDF failure modes with tests

Add encryption detection and PyPI packaging

Improve untagged accuracy ordering with stream-first fallback

Refactor extraction modes and add memory regression guards

Merge pull request #6 from Lulzx/dev

docs: fix README inaccuracies in features, reading order, and project structure

fix: use wrapping arithmetic for octal escape parsing; fix README writer example

fix: plug memory leaks and negative-cast panics found in code review

Merge pull request #5 from Lulzx/dev

fix: heap-allocate StructElements to prevent dangling pointers; expand tests

fix: address PR #5 review comments

Merge pull request #3 from trvon/windows-support

Merge pull request #4 from phenomen/wapi-markdown