fix: parse markdown buffer with markdown parser and get inlines as children

[jmbuhr]

This fixes interactions with other plugins by not accidentally attaching a wrong parser to markdown buffers (that would get re-used by other plugins when they require the parser).

It should also safeguard against future bugs, because previously we would attach the "markdown_inline" parser directly to the buffer, which would also try to parse all the parts that are not plain markdown (like yaml headers, html and code chunks) as inline markdown.

This does change the behavior slightly, but I'd say makes it more expected. Previously, because the whole buffer would be treated as if it where only "markdown_inline", it would also detect images with the markdown syntax in parts that are not markdown, so this is no longer happening. e.g. if you had a code chunk with a comment that contained some markdown image syntax.

# Hello world

```{python}
# this is a comment with ![](./some/image.png)
# this will, rightfully, not be rendered anymore
print('hello world')
```

But this will: ![](./some/image.png)

[3rd]

Amazing work, thank you so much!