- Encode bigint bytes by bit-shift mask instead of the broken accumulator
- Decode the bigint length prefix and payload bytes unsigned
- Encode negative bigint as dedicated type 8 with magnitude payload
- Un-skip the multi-byte and heterogeneous bigint cases, add focused cases
- Correct the int64 decode multiplier to 2147483648
- Encode negative int64 as dedicated type 5 with magnitude payload
- Shift float/bigint/long-string/short-string type codes to 6/7/9/length+9
- Un-skip the int64 round-trip case and add negative int64 focused cases
