Implement an AVX2 alg based on Wojciech Muła & Daniel Lemire 2017

The paper this algorithm is taken from can be found via
DOI 10.1145/3132709
arXiv:1704.00605
or via arXiv.org: https://arxiv.org/abs/1704.00605

Author: dequbed

Cargo.toml

@@ -27,6 +27,8 @@ rstest_reuse = "0.1.3"
 default = ["std"]
 alloc = []
 std = []
+unsafe = []
+avx2 = ["unsafe"]
 
 [profile.bench]
 # Useful for better disassembly when using `perf record` and `perf annotate`