Introduce murmur3_32_of_slices

[mge-fbe-com]

Add murmur3_32_of_slices that reads directly from multiple byte buffers.

According to bench it's usually slightly faster on x86_64 (Broadwell)

test bench_32_slice  ... bench:           5.71 ns/iter (+/- 0.79) = 11200 MB/s
test bench_32_slices ... bench:           5.56 ns/iter (+/- 0.82) = 11200 MB/s

6% faster on AMD EPYC 7D13

test bench_32_slice  ... bench:           8.47 ns/iter (+/- 0.05) = 7000 MB/s
test bench_32_slices ... bench:           7.94 ns/iter (+/- 0.03) = 8000 MB/s

18% on aarch64 (apple M4):

test bench_32_slice      ... bench:           4.14 ns/iter (+/- 0.39) = 14000 MB/s
test bench_32_slices     ... bench:           3.37 ns/iter (+/- 0.03) = 18666 MB/s