Add a "ZigZag" encoder for signed integers that stores the sign bit as the least significant bit. Two's complement is using the most significant bit resulting in a large bit difference for -1 vs 0.

With ZigZag, Simple-8b encoder does then not have to know about signed integers and can treat everything as unsigned.

For 32bit integers ZigZag encoding would look like this:
zigzag = (n << 1) ^ (n >> 31)