Export binary content document nodes as JS ArrayBuffer objects directly

This removes the need for explicit conversion and some other hacks to work
around the lack of access to opaque QByteArray variants.