without this, the code fence parser was getting thrown off if you didn't
have the pydot extension loaded, which was thwarting backwards
compatibility. this makes the pydot bits look like an attribute to the
vanilla parser, so at least then the vanilla markdown renders as
intended