Consider using alternative markdown parser

[ekropotin]

Using another CM parser, e.g. https://github.com/pulldown-cmark/pulldown-cmark, poetically can give significant performance boost.
There some benchmarking was done in this thread: kivikakk/comrak#25

One thing that turned me off initially from using pulldown-cmark is that it doesnt seem provide the way to differentiate ATX-style headers from setex-style headers. It makes implementation of some rules, such as MD003, impossible.