security: if chat Markdown is added, require marked + DOMPurify allowlist

[krwg]

Summary

Preventive hardening note: chat is currently safe (createElement / textContent in chat-message-content.js). marked + DOMPurify already exist for release notes. If chat Markdown ships later, it must reuse sanitize with a strict allowlist — never ad-hoc innerHTML from peer text.

Requirements when the feature lands

• marked.parse → DOMPurify.sanitize with explicit ALLOWED_TAGS / URI rules
• No unrestricted <img> / javascript: hrefs
• Prefer extracting a shared helper next to release-markdown.js rather than a second sanitizer in chat

Acceptance (for this tracking issue)

• Constraint captured in SECURITY or architecture docs now
• When Markdown chat is implemented, PR checklist includes sanitize allowlist review