The reason for the switch is mostly related to error handling (and me
being unable to unterstand how nom does it). Using byteorder and a few
handwritten helper functions for parsing makes error handling much
easier. It also allows me to return more helpful error messages.