This is mostly because I am not 100% certain about how stable our own
optimized version is. While it does perform a lot better, there is
significant risk that things might break in production. To avoid
that, I am replacing it with the upstream version which is much more
heavily tested.
This replaces DOMParser with htmlparser2 which is much, much faster.
How much faster? 80%. This new implementation can parse at 50mb/s
which is insane! The old one could only do 5-10mb/s
We still haven't gotten rid of the DOMParser though since HTML-to-MD
conversion still needs it. This will be done soon though by using `dr-sax`.
This uses a custom implementation of htmlparser2 instead of the default
one which is 50% faster.
