[Performance] Possible optimizations

[jpmoutinho]

https://github.com/quantumlib/qsim

To speed up the simulator, we use gate fusion [2] [3], single precision arithmetic, AVX/FMA instructions for vectorization and OpenMP for multi-threading.

I suspect this gate fusion is similar to what is already implemented in PyQTorch for 1Q gates, but maybe there are some more general ideas to explore there.

[dominikandreasseitz]

@jpmoutinho yes, gate-fusion is what we do in the pyq backebd in qadebce