Contrary to what was implemented it seems that UDTs and tuples are
encoded as a single bytestring and their internal structure is that of
consecutive CQL values. Additionally--as pointed out in #5--collections
may contain UDTs in addition to native types.