Use scala-offheap for performance

Not sure how applicable this is to Jawn, but should we look at using https://github.com/densh/scala-offheap to both improve the performance (but more importantly the latency/memory characteristics) of the parser?
