Allow running bandwidth benchmarks with different kernels

[giordano]

Follow up from https://github.com/giordano/julia-on-fugaku/issues/1: it'd be nice to allow running bwscaling (and necessarily also bwbench) with kernels other than sdxapy. This would allow comparing results with https://github.com/RRZE-HPC/TheBandwidthBenchmark, which basically reports scaling all other kernels but sdaxpy: init, sum, copy, update, triad. Example: Fujitsu FX700 S1 M4 C48.

[carstenbauer]

Wrong repository, see https://github.com/JuliaPerf/BandwidthBenchmark.jl/issues/9.