How difficult would it be to also expose the [SAX interface](http://rapidjson.org/md_doc_sax.html) so we can process files that are much larger than can fit in RAM? I am currently using `ijson` which has a `yajl` backend, but it's [about 5x slower than RapidJSON](https://chadaustin.me/2013/01/json-parser-benchmarking/).