One of the main differences is this one uses lookup tables for faster decoding of Huffman compressed data as opposed to using a tree (which is what the stdlib one uses). It also offers some additional encoding options for users that want more control of header field indexing.
Yup! That documentation is too daunting IMO but it's complete and not many steps, despite all the words. Especially if you're a Go programmer already. (just go get the code review tool)
I'd start with filing a bug against Go to track the overall process.