regexp tests: make some benchmarks more meaningful
The first two "Literal" tests were silly.
Also fix a naming typo in a couple of tests.

R=rsc
CC=golang-dev
https://golang.org/cl/4127045
diff --git a/src/pkg/regexp/all_test.go b/src/pkg/regexp/all_test.go
index aed7330..c7ee4c8 100644
--- a/src/pkg/regexp/all_test.go
+++ b/src/pkg/regexp/all_test.go
@@ -316,9 +316,9 @@
 }
 
 func BenchmarkLiteral(b *testing.B) {
-	x := strings.Repeat("x", 50)
+	x := strings.Repeat("x", 50) + "y"
 	b.StopTimer()
-	re := MustCompile(x)
+	re := MustCompile("y")
 	b.StartTimer()
 	for i := 0; i < b.N; i++ {
 		if !re.MatchString(x) {
@@ -329,9 +329,9 @@
 }
 
 func BenchmarkNotLiteral(b *testing.B) {
-	x := strings.Repeat("x", 49)
+	x := strings.Repeat("x", 50) + "y"
 	b.StopTimer()
-	re := MustCompile("^" + x)
+	re := MustCompile(".y")
 	b.StartTimer()
 	for i := 0; i < b.N; i++ {
 		if !re.MatchString(x) {