Code cell in blockquote does not execute

[rowanc1]

The following nested code-cell fails execution.

```{code-cell} python
hello = "hello"
there = "there"
phrase = f"{hello}, {there}!"
print(phrase)
```
some markdown code example

:tags: remove-input
print("This will show output with no input!")

> ```{code-cell} python
> :tags: remove-input
> print("This will show output with no input!")
> ```

the above block

This is the AST:

The failure:

That is the markdown cell that we are trying to execute in python.

The selection here:
https://github.com/executablebooks/mystmd/blob/bug/code-block/packages/myst-execute/src/execute.ts#L282

it does not take into account nested blocks, and/or special blocks that should only contain a code/output pair.
Instead there is a block that contains a code-cell.

The next error is here:
https://github.com/executablebooks/mystmd/blob/bug/code-block/packages/myst-execute/src/execute.ts#L120

Which I think needs an executable flag on it.

Regardless, needs some more test cases on these nested select cases, and/or being more specific about the block we are filtering for.

[agoose77]

Ah, it looks like the selector is wrong
block:has(code[executable=true]):has(output),inlineExpression

I hadn't thought about executable nodes being nested inside quotes. I didn't think we wanted to support that kind of usage, but I have not personally given it much thought (and I reckon I'd be in favour of it).

The issue is that our version of unist-util-select (and in turn, unist) doesn't let us tighten the descendent selector in has(child) to `has(> child), which is unfortunate.

I'll give this some cycles on Monday, when I'm back in the office, but anyone else is welcome to tackle it!

[rowanc1]

I think there are some quick tightening we can do, including labelling the block on the way through. The other block is actually a markdown cell, which happens to contain a code-cell. Not allowed in Jupyter Notebooks, but probably fine to allow here, and I kinda like that it should work.

The other selector should at least look for an executable code cell. That would have made it "work" but the one cell would be executed twice.

I think the better plan is to use the data: type here or kind: code-cell to make the selection of blocks much clearer and less based on the structure. I think we talked around parts of this before, but forget what we decided!

[rowanc1]

@agoose77 I have put together a fix in #1136 by changing block.data.type --> block.kind, which is going to specifically select those blocks. It does seem to work and pass various tests, I want @fwkoch's take on this, and I haven't tested with thebe.

I think it is good for the cell kind to be a known attribute of the block.