I very much like, what you have done with the code. I agree, that it is nice to be able to declare pandoc extensions.
On the other hand, I disapproved at taking out pypandoc at first. However, after taking a closer look at the source of pypandoc, I came to realize, that in this case we do not hugely benefit by using an abstraction layer.
Your code works nicely with my setup with the exception that you removed the change of the cwd. However, as pandoc_reader is still a very young project, I think we can afford to break compatibility.
Thank you very much for your contribution!
* Add PANDOC_EXTENSIONS configuration variable, allowing one to
enable or disable Pandoc's markdown extensions individually.
* Remove dependency on pypandoc.
* Don't change the working directory.
* More efficient metadata extraction.
