Describe the enhancement requested

According to previous observations, it seems the AVX512 accelerations of BYTE_STREAM_SPLIT perform equal or worse then their AVX2 counterparts.

Besides, the SSE2 and AVX2 accelerations, already performing at 5-10 GB/s or more, are amply fast enough.

We could therefore simply remove the AVX512 accelerations.

Component(s)

Benchmarking, C++, Parquet