language-mime metadata for displaying text/plain output

[takluyver]

From the discussion on #356, we want to reuse the syntax highlighting we already have for code input to support code output. Our idea is that this would be displayed as a display_data or execute_result output, with text/plain mimetype, and language-mime metadata (name still open to discussion) identifying the highlighting to use. This degrades nicely for clients that don't recognise it: they will display the code unhighlighted.

• Use this metadata in nbconvert to highlight with pygments
• Use this metadata in the notebook to highlight with codemirror
• Add functionality to kernels to send display output with this metadata
• Resolve the discrepancy between mimetypes for R source code (pygments/codemirror)

[mgeier]

Great! This would allow implementing a magic command similar to %load, but showing the content of a file in a syntax-highlighted output cell instead of an input (code) cell. See https://stackoverflow.com/q/19197931/.