Generate MyST AST from cell execution

[choldgraf]

The executable plugins documentation for directives shows how you can return MyST AST by printing to stdout in order to generate different types of MyST outputs.

Users should be able to do the same thing via {code-cell} directives and Jupyter Notebook cells. This would allow users to more flexibly generate MyST AST without requiring any kernel-specific libraries (since you'd just be using stdout).

Implementation ideas

Doing it with tags

One approach could be to use cell tags to tell MyST when to capture a cell's output and treat it as MyST AST.
Here's an example of what this could look like:

```{code-cell} python
:tags: output-myst-ast
ast = {
      "type": "list",
      "ordered": false,
      "spread": false,
      "children": [
        {
          "type": "listItem",
          "spread": True,
          "children": [
            {
              "type": "strong",
              "children": [
                {
                  "type": "text",
                  "value": "Bolded"
                }
              ]
            },
            {
              "type": "text",
              "value": " list item"
            }
          ]
        }
      ]
    }
print(ast)
```

[agoose77]

@choldgraf I never had the chance to engage on this -- I don't think we need to use a custom tag. e.g. for IPython:

display({
  "application/vnd.mystmd.ast+json;version=1": {
    "type": "root",
    "children": [
      {
        "type": "paragraph",
        "children": [
          {
            "type": "emphasis",
            "children": [
              {
                "type": "text",
                "value": "Hello"
              }
            ]
          }
        ]
      }
    ]
  }
}, raw=True)

We should definitely support this in addition to parsing MyST Markdown directly, but the technical difficulty of doing both is identical.

[choldgraf]

My main concern about not using tags and using a function instead is:

It depends on there being a display function, which exists in Python, but not for all kernels. It means that each language would need to write their own version of display, right? I was trying to find an implementation that would require no language-specific development.

[agoose77]

@choldgraf I think all kernels need to implement the display protocol, so if we restrict ourselves to R, C++, and Python, it's true that there's a default publication API IIRC.

[choldgraf]

Ah cool - you know more about this one than I do :-) My main concern is that we don't require a lot of language-specific development for each language that we want to support MyST outputs. But as long as that's simple, I'm +1 on whatever approach is easiest and has the best UX.