Consider using a proper Markdown parser to remove non-textual content

[Eilon]

Related: #62

If we use a proper Markdown parser it would be great to take some Markdown like this:

> Hey brothers, check out this radical *code*:
> ```javascript
> var x = document.foo(...);
> var bananas = monkey.pole();
> ```
> And also what if we frobbed the `bananas` variable and instead used a `fruit` syntax model?
> # Alternative solution
> What if instead of fruit, we supported the `MSFT.Botany` package?

And trim it down to something like this:

> Hey brothers, check out this radical code:
> And also what if we frobbed the  variable and instead used a  syntax model?
> Alternative solution
> What if instead of fruit, we supported the  package?

Note that all code blocks between backticks, all formatting marks (asterisk, underline, etc.), header marks (after hashes) are removed.

And then run the textual analysis on only those sentences.