Replaced RANS0 and RANS1 codec identifiers with a single RANS
codec. Internally we still call them RANS0 and RANS1 methods so it
can auto-tune to order-0 or order-1 in the statistics, but before
writing out we force both to be RANS/RANS0 identifier.
The rANS implementation has therefore added an extra byte (order) to
the data stream. I also tweaked the efficiency of decoding a bit more
so it's now faster.