Script that checks DSL examples in documentation are up to date

[cdleary]

Probably good do this in two steps: introduce language blocks for DSL examples that indicate which ones are "full" and testable (vs snippets), and then make a script that scrapes all the full ones out of the docs and parses/typechecks/tests them.

[cdleary]

I spent some time poking around inside Python-Markdown (which mkdocs uses) to see if hooking into existing structured parsing in there would be smarter than just regexp-ing our .md files. The way that the fenced code extension works, it seems to effectively match out of the character stream with a regexp, and it seems to want the fenced code blocks just at the document root level with the default extension (per details https://python-markdown.github.io/extensions/fenced_code_blocks/ ) so probably not so bad to just implement the simple thing ourselves: https://github.com/Python-Markdown/markdown/blob/c0b1dc5d315624799ac1ba380eced9838e7f4bca/markdown/extensions/fenced_code.py#L43