Avoid string copies.

[pgavlin]

Instead of copying strings to a byte slice in the implementation of Iter et. al., use unsafe to get at the strings' bytes. These byte slices are never mutated, so this is safe.

[pgavlin]

benchstat output:

aho-corasick ❯ benchstat base.txt diff.txt 
goos: darwin
goarch: arm64
pkg: github.com/petar-dambovaliev/aho-corasick
cpu: Apple M4 Max
                                            │   base.txt   │              diff.txt               │
                                            │    sec/op    │   sec/op     vs base                │
Stdlib_StringsReplaceAll/No_matches-16        58.54µ ±  1%   58.49µ ± 1%        ~ (p=0.796 n=10)
Stdlib_StringsReplaceAll/Matches-16           52.24µ ±  1%   53.14µ ± 1%   +1.72% (p=0.002 n=10)
Stdlib_AhoCorasickReplaceAll/No_matches-16    46.15µ ±  2%   43.27µ ± 1%   -6.26% (p=0.000 n=10)
Stdlib_AhoCorasickReplaceAll/Matches-16       55.84µ ±  1%   49.72µ ± 1%  -10.96% (p=0.000 n=10)
AhoCorasick_ReplaceAllDFA-16                  53.12µ ± 18%   47.33µ ± 1%  -10.89% (p=0.000 n=10)
AhoCorasick_ReplaceAllNFA-16                  58.80µ ±  0%   53.02µ ± 1%   -9.84% (p=0.000 n=10)
AhoCorasick_LeftmostInsensitiveWholeWord-16   49.36µ ±  1%   46.24µ ± 1%   -6.32% (p=0.000 n=10)
geomean                                       53.26µ         49.95µ        -6.21%
                                            │   base.txt   │                diff.txt                │
                                            │     B/op     │     B/op      vs base                  │
Stdlib_StringsReplaceAll/No_matches-16        112.0Ki ± 0%   112.0Ki ± 0%        ~ (p=1.000 n=10)
Stdlib_StringsReplaceAll/Matches-16           112.0Ki ± 0%   112.0Ki ± 0%        ~ (p=1.000 n=10) ¹
Stdlib_AhoCorasickReplaceAll/No_matches-16    57456.0 ± 0%     112.0 ± 0%  -99.81% (p=0.000 n=10)
Stdlib_AhoCorasickReplaceAll/Matches-16       180.3Ki ± 0%   124.2Ki ± 0%  -31.09% (p=0.000 n=10)
AhoCorasick_ReplaceAllDFA-16                  181.7Ki ± 0%   125.5Ki ± 0%  -30.90% (p=0.000 n=10)
AhoCorasick_ReplaceAllNFA-16                  181.5Ki ± 0%   125.4Ki ± 0%  -30.92% (p=0.000 n=10)
AhoCorasick_LeftmostInsensitiveWholeWord-16   58240.0 ± 0%     872.0 ± 0%  -98.50% (p=0.000 n=10)
geomean                                       113.2Ki        21.73Ki       -80.81%
¹ all samples are equal
                                            │  base.txt  │               diff.txt               │
                                            │ allocs/op  │ allocs/op   vs base                  │
Stdlib_StringsReplaceAll/No_matches-16        3.000 ± 0%   3.000 ± 0%        ~ (p=1.000 n=10) ¹
Stdlib_StringsReplaceAll/Matches-16           3.000 ± 0%   3.000 ± 0%        ~ (p=1.000 n=10) ¹
Stdlib_AhoCorasickReplaceAll/No_matches-16    3.000 ± 0%   2.000 ± 0%  -33.33% (p=0.000 n=10)
Stdlib_AhoCorasickReplaceAll/Matches-16       23.00 ± 0%   22.00 ± 0%   -4.35% (p=0.000 n=10)
AhoCorasick_ReplaceAllDFA-16                  81.00 ± 0%   77.00 ± 0%   -4.94% (p=0.000 n=10)
AhoCorasick_ReplaceAllNFA-16                  64.00 ± 0%   60.00 ± 0%   -6.25% (p=0.000 n=10)
AhoCorasick_LeftmostInsensitiveWholeWord-16   21.00 ± 0%   19.00 ± 0%   -9.52% (p=0.000 n=10)
geomean                                       13.14        11.95        -9.07%
¹ all samples are equal

[petar-dambovaliev]

Hello. Thanks for the PR. I am hesitant in using unsafe here. I will need to think about it.