Sparse byte changes can reduce waveform encoding CPU

Tested exact byte-level sparse innovations: encode positive gaps between changed byte positions, the nonzero XOR bytes, and a per-event terminator, then apply zstd. On additional sampled wide-struct blocks it reduced storage and both encode/decode CPU versus optimized full-width XOR plus zstd-9, including a separate control that drops zero X/Z planes. A bitmap variant was smaller but slower to decode. An earlier XOR wrapper also had avoidable per-value allocation work, so benchmark against a contiguous word-wise implementation before crediting the new representation. Results remain design- and sample-specific; related changed-field approaches are established prior art.