getRunnableContentFromString` doesn't behave well when there are multiple code blocks. It currently will match all of the text between the first code block and the last one, and when the user runs it this will either error or do some unexpected behavior.

Instead we should iterate over the content returned, and when there's a full code block, we then can add the run actions. That way we allow for multiple code actions and also don't end making folks run the wrong code.

We could only return the runnable content from the last code block, this would be the easiest fix, but we should avoid this as ideally folks should be able to run any code block in the response.