Add prosemirror pkg to handle conversion from and to other formats

[codenem]

TODO

• MD to prosemirror JSON cli
• Does the frontend tiptap support all previous docs
• Unwanted page breaks

---

Summary by cubic

Add pkg/prosemirror to parse ProseMirror/Tiptap JSON and render safe HTML for PDFs. Add a Markdown→ProseMirror converter and CLI tools for content migration; enable h1–h6 in the editor; switch pkg/docgen to ProseMirror, and fix unwanted PDF page breaks.

• New Features

  • pkg/prosemirror: Node/Mark structs and RenderHTML(Node) that escapes HTML and adds language-* on code blocks. Supports paragraphs, headings (h1–h6), blockquote, horizontal rules, hard breaks, images, bullet/ordered lists, and tables (cells/headers with colspan/rowspan; colwidth → min-width). Marks: bold, italic, underline, strike, code, link (href/target/rel/class/title).
  • Markdown import: ParseMarkdown([]byte) builds ProseMirror JSON using github.com/yuin/goldmark for consistent lists, tables, code, and inline marks.
  • CLI: proboctl md-to-prosemirror (stdin → JSON), probo documentversion update-content to push JSON via the API, and migrate-document-versions-markdown to rewrite document_versions.content from Markdown to ProseMirror JSON.
  • Rich editor: enable headings up to h6, add Heading 4 to menus, and update CSS sizes for h1–h6.
  • Docs: added contrib/claude/go-style.md and linked it from AGENTS.md.

• Refactors

  • pkg/docgen: render from ProseMirror JSON → HTML; template now uses .BodyHTML, styles blockquotes, and removes H2-based page-break CSS. On parse/render failure, fallback to a single escaped <p> with the raw input. Tests updated to expect ProseMirror HTML structure.
  • Dependencies: removed go.abhg.dev/goldmark/mermaid; upgraded github.com/yuin/goldmark to v1.8.2 and use it only in the Markdown→ProseMirror converter.

^{Written for commit 3815eec. Summary will update on new commits.}

[gearnode]

json.RawMessage?