remove StartTimer/StopTimer from the bench loop

It can completely mess up the benchmark numbers for init functions that
were too small. Moreover, it could make the -benchtime estimates be way
off. For example, 'benchinit cmd/go' was taking over a minute to run the
benchmark, instead of the expected ~1s.

The benchtime estimate being off is likely the upstream issue
golang/go#27217.

The fact that StartTimer and StopTimer are expensive is being tracked in
golang/go#20875.

main.go

@@ -334,6 +334,11 @@ type toZero struct {
 	Offset, ZeroSize int64
 }
 
+// Don't use StartTimer and StopTimer, as they call ReadMemStats, which is way
+// too expensive compared to most init functions.
+// For example, on 'benchinit cmd/go', ReadMemStats was taking up over 90% of
+// the cpu time, and throwing off all the numbers.
+
 var benchTmpl = template.Must(template.New("").Parse(`
 // Code generated by benchinit. DO NOT EDIT.
 
@@ -349,11 +354,8 @@ func BenchmarkInit(b *testing.B) {
 	b.ReportAllocs()
 
 	for i := 0; i < b.N; i++ {
-		b.StartTimer()
-		_init()
-		b.StopTimer()
-
 		deinit() // get ready to run init again
+		_init()
 	}
 }