accelerate by 5% on simple string and 2% on complex cases

Acceleration is only visible on Intel i7 10th gen, not visible on
Raspberry Pi 3 (but there's no regression either).

$ go install github.com/maruel/pat/cmd/...@latest
$ ba
go test -bench . -benchtime 100ms -count 5 -run ^$ -cpu 1
Checking out origin/main
go test -bench . -benchtime 100ms -count 5 -run ^$ -cpu 1
Checking out accel
go test -bench . -benchtime 100ms -count 5 -run ^$ -cpu 1
Checking out origin/main
go test -bench . -benchtime 100ms -count 5 -run ^$ -cpu 1
Checking out accel
name                       old time/op    new time/op    delta
LessDigitsTwoGroupsNative    3.63ns ± 0%    3.64ns ± 1%    ~     (p=0.440 n=8+10)
LessDigitsTwoGroups          37.9ns ± 1%    37.1ns ± 2%  -2.11%  (p=0.000 n=10+10)
LessStringOnly               8.09ns ± 1%    7.69ns ± 0%  -4.94%  (p=0.000 n=10+10)
LessDigitsOnly               17.8ns ± 0%    17.8ns ± 2%    ~     (p=0.586 n=8+10)
Less10Blocks                  198ns ± 1%     194ns ± 1%  -1.93%  (p=0.000 n=10+10)

name                       old alloc/op   new alloc/op   delta
LessDigitsTwoGroupsNative     0.00B          0.00B         ~     (all equal)
LessDigitsTwoGroups           0.00B          0.00B         ~     (all equal)
LessStringOnly                0.00B          0.00B         ~     (all equal)
LessDigitsOnly                0.00B          0.00B         ~     (all equal)
Less10Blocks                  0.00B          0.00B         ~     (all equal)

name                       old allocs/op  new allocs/op  delta
LessDigitsTwoGroupsNative      0.00           0.00         ~     (all equal)
LessDigitsTwoGroups            0.00           0.00         ~     (all equal)
LessStringOnly                 0.00           0.00         ~     (all equal)
LessDigitsOnly                 0.00           0.00         ~     (all equal)
Less10Blocks                   0.00           0.00         ~     (all equal)

README.md

@@ -19,11 +19,11 @@ goos: linux
 goarch: amd64
 pkg: github.com/maruel/natural
 cpu: Intel(R) Core(TM) i7-10700 CPU @ 2.90GHz
-BenchmarkLessDigitsTwoGroupsNative 329085624     3.628 ns/op   0 B/op   0 allocs/op
-BenchmarkLessDigitsTwoGroups        31792579    38.07 ns/op    0 B/op   0 allocs/op
-BenchmarkLessStringOnly            147547137     8.127 ns/op   0 B/op   0 allocs/op
-BenchmarkLessDigitsOnly             65866989    17.92 ns/op    0 B/op   0 allocs/op
-BenchmarkLess10Blocks                5997386   198.3 ns/op     0 B/op   0 allocs/op
+BenchmarkLessDigitsTwoGroupsNative 331287298     3.597 ns/op   0 B/op   0 allocs/op
+BenchmarkLessDigitsTwoGroups        32479050    36.55 ns/op    0 B/op   0 allocs/op
+BenchmarkLessStringOnly            157775884     7.603 ns/op   0 B/op   0 allocs/op
+BenchmarkLessDigitsOnly             69210796    17.52 ns/op    0 B/op   0 allocs/op
+BenchmarkLess10Blocks                6331066   190.8 ns/op     0 B/op   0 allocs/op
 ```
 
 On a Raspberry Pi 3:
@@ -33,11 +33,11 @@ $ go test -bench=. -cpu 1
 goos: linux
 goarch: arm
 pkg: github.com/maruel/natural
-BenchmarkLessDigitsTwoGroupsNative  13044535    85.86 ns/op    0 B/op   0 allocs/op
-BenchmarkLessDigitsTwoGroups         1576779   751.7 ns/op     0 B/op   0 allocs/op
-BenchmarkLessStringOnly              8470698   141.5 ns/op     0 B/op   0 allocs/op
-BenchmarkLessDigitsOnly              3674454   326.4 ns/op     0 B/op   0 allocs/op
-BenchmarkLess10Blocks                 314845  3821 ns/op       0 B/op   0 allocs/op
+BenchmarkLessDigitsTwoGroupsNative  14181789    86.57 ns/op    0 B/op   0 allocs/op
+BenchmarkLessDigitsTwoGroups         1600195   748.9 ns/op     0 B/op   0 allocs/op
+BenchmarkLessStringOnly              8286034   142.3 ns/op     0 B/op   0 allocs/op
+BenchmarkLessDigitsOnly              3653055   331.4 ns/op     0 B/op   0 allocs/op
+BenchmarkLess10Blocks                 310687  3838 ns/op       0 B/op   0 allocs/op
 ```
 
 Coverage:
@@ -48,4 +48,3 @@ PASS
 coverage: 100.0% of statements
 ok     github.com/maruel/natural       0.012s
 ```
-

natsort.go

@@ -21,13 +21,13 @@ import (
 // This function does no memory allocation.
 func Less(a, b string) bool {
 	for {
-		if a == b {
-			return false
-		}
 		if p := commonPrefix(a, b); p != 0 {
 			a = a[p:]
 			b = b[p:]
 		}
+		if len(a) == 0 {
+			return len(b) != 0
+		}
 		if ia := digits(a); ia > 0 {
 			if ib := digits(b); ib > 0 {
 				// Both sides have digits.