Add mermaid diagram sanitization to wiki push step

[Copilot]

GitHub's wiki mermaid renderer rejects two syntax patterns the AI model occasionally generates, causing "Unable to render rich display" errors. This adds a post-processing sanitization step that fixes these patterns before committing wiki pages.

Changes

• New "Sanitize Mermaid diagrams" step in the push-wiki safe-output job, running between "Write wiki pages" and "Commit and push"
  • Strips backtick markdown-string syntax from node labels (A["`text`"] → A["text"])
  • Handles embedded \n escape sequences within backtick labels (A["`prefix`\nextra"] → A["prefix extra"])
  • Replaces any remaining \n escape sequences in labels with spaces
  • Reports which files were modified

Example

Before (causes lexer error):

flowchart LR
  core["`@react-term/core`\nCellGrid, Buffer, VTParser"] --> web

Loading

After sanitization:

flowchart LR
  core["@react-term/core CellGrid, Buffer, VTParser"] --> web

Loading

This acts as a reliable safety net regardless of what the model generates, since the existing self-review checklist instructions alone are insufficient to prevent the issue.

---

📱 Kick off Copilot coding agent tasks wherever you are with GitHub Mobile, available on iOS and Android.