feat: replace regex README parser with markdown-it-py AST parser

Introduce parse_readme() which uses MarkdownIt to build a full AST
instead of line-by-line regex matching. The function splits the document
at the thematic break, groups nodes by h2 heading, extracts category
descriptions from leading italic paragraphs, and separates the
Categories, Resources, and Contributing sections cleanly.

Add markdown-it-py==4.0.0 (+ mdurl) as a runtime dependency to support
the new parser.

Tests cover section counts, names, slugs, descriptions, content
presence, boundary conditions (no separator, no description), and mixed
description markup.

Co-Authored-By: Claude <[email protected]>