Add an `eval_rst` directive

[choldgraf]

In #163 we identified that there may be circumstances where the author still needs to write rST with MyST. For example, if a person uses autodoc and wants their documentation written in MyST, but has docstrings written in rST.

In that case, we could include a directive like recommonmark's eval_rst that will call the rST parser on its content.

Another possible approach (maybe this is unrealistic?) is to allow for a flag in all directives used by MyST-parser that could tell the directive to parse the content with the rST parser instead of the MyST parser.

[chrisjsewell]

Another possible approach (maybe this is unrealistic?) is to allow for a flag in all directives used by MyST-parser that could tell the directive to parse the content with the rST parser instead of the MyST parser.

From here #163 (comment), it looks like sphinx does some weird stuff with the auto directives, actually rewriting the directive in rST format before parsing. In this case that approach would not work.

[choldgraf]

Just a note that it looks like this is where recommonmark does this: https://github.com/readthedocs/recommonmark/blob/c4d575bd7bd5dd992cc3e2bed9dce625f4622a4d/recommonmark/transform.py#L246

It looks like they pass the contents to the RST state parser, which I think we have access to in MyST? We're using a Markdown "mock" parser, so hopefully the rST parser is available as well

[asmeurer]

I was thinking, if it isn't hard, it might be useful to also provide the inverse directive, like .. eval_markdown:: for RST. That way things that currently have to be written in RST like #163 can still include Markdown segments.