Markdown import parser upgrade

[TomMaSS]

Summary

Replace the heuristic markdown import parser with markdown-it-py AST parser for more reliable section splitting.

Motivation

Current parser uses regex-based heading detection which can misparse edge cases (headings in code blocks, non-standard markdown).

Technical Approach

• Replace heuristic parser in prd_import_markdown with markdown-it-py
• Parse to AST, walk tree to find heading nodes
• Split content at h2 boundaries (configurable heading level)
• Preserve code blocks, lists, and other markdown structures correctly

Key Files

• mcp_server/tools/import_tools.py — prd_import_markdown implementation
• requirements.txt — add markdown-it-py

Acceptance Criteria

• Parser uses markdown-it-py AST
• Headings in code blocks not treated as section boundaries
• Existing import behavior preserved for well-formed markdown
• Unit tests for edge cases