common/bitutil: optimize ANDBytes

[cuiweixie]

goos: darwin
goarch: arm64
pkg: github.com/ethereum/go-ethereum/common/bitutil
cpu: Apple M1 Pro
              │    old.txt    │               new.txt               │
              │    sec/op     │   sec/op     vs base                │
FastAND1KB-10    414.00n ± 0%   20.37n ± 2%  -95.08% (p=0.000 n=10)
FastAND2KB-10    822.70n ± 2%   35.48n ± 3%  -95.69% (p=0.000 n=10)
FastAND4KB-10   1710.00n ± 8%   65.80n ± 0%  -96.15% (p=0.000 n=10)
geomean           835.1n        36.23n       -95.66%

              │   old.txt    │               new.txt               │
              │     B/op     │    B/op     vs base                 │
FastAND1KB-10   0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
FastAND2KB-10   0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
FastAND4KB-10   0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
geomean                    ²               +0.00%                ²
¹ all samples are equal
² summaries must be >0 to compute geomean

              │   old.txt    │               new.txt               │
              │  allocs/op   │ allocs/op   vs base                 │
FastAND1KB-10   0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
FastAND2KB-10   0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
FastAND4KB-10   0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
geomean                    ²               +0.00%                ²
¹ all samples are equal
² summaries must be >0 to compute geomean