Add MarkdownToPDF and CreateInvoiceFromStructuredData samples

[datalogics-robb]

Rewrites the two new DocumentConversion samples from apdfl-csharp-dotnet-samples (PRs datalogics/apdfl-csharp-dotnet-samples#106 and #109, ported to develop-21 in #114) using the Modern C++ interface.

MarkdownToPDF

Converts a practical Markdown subset directly to a newly created, tagged PDF: ATX/setext headings, emphasis with CommonMark-style flanking rules, inline code with backdrop, strikethrough, links (inline/reference/autolink/bare, drawn as working LinkAnnotation + URIAction), ordered/unordered/task lists, fenced code blocks with language title rows and preserved indentation, blockquotes, pipe tables with per-column alignment, horizontal rules, and an HTML-lite subset normalized before parsing. Text is laid out with Font::measure_text_width; CJK wraps per character and CJK/Cyrillic/Greek fall back to installed fonts via Font::is_text_representable. --self-test runs the parser tests without writing a PDF.

CreateInvoiceFromStructuredData

Builds a tagged PDF invoice from JSON metadata, CSV line items, a JSON style sheet, and a local PNG logo (tagged Figure with alt text). The line-item table repeats its header on continuation pages; footers/rules/boxes are emitted as artifacts. When enabled in the metadata, the document is secured with Document::secure (AES-256, owner password) so it can be viewed and printed but not edited. JSON/CSV parsing is a small self-contained parser, keeping APDFL the only dependency.

Notes

• Both samples build the structure tree with StructTreeRoot / StructElement::add_marked_content_ref — the Modern C++ API assigns MCIDs and maintains the parent tree automatically, so the tagging code is much simpler than the .NET equivalents.
• Verified on macOS against ModernCppSDK 21.1.0: both samples build clean, self-tests pass, all bundled sample documents convert, the structure trees were validated with the ListStructTree sample, and the rendered pages were visually inspected.
• Character classification in the Markdown parser is deliberately ASCII-explicit rather than <cctype>: APDFL's Library init changes the process locale, which otherwise misclassifies UTF-8 continuation bytes (this bit during development).
• Also adds the missing CreateDocFromWebPage entry to the DocumentConversion README and corrects the sample count in CLAUDE.md.
• Batch directory conversion (--recursive) from the .NET MarkdownToPDF is intentionally not carried over.

🤖 Generated with Claude Code