runtime: add goroutine creation benchmark

Current results on 2 core darwin/amd64:
BenchmarkGoroutineChain		351 ns/op
BenchmarkGoroutineChain-2	3840 ns/op
BenchmarkGoroutineChain-4	4040 ns/op
BenchmarkConcGoroutineChain	350 ns/op
BenchmarkConcGoroutineChain-2	875 ns/op
BenchmarkConcGoroutineChain-4	2027 ns/op

R=r, rsc
CC=golang-dev
https://golang.org/cl/6332054

src/pkg/runtime/proc_test.go

@@ -123,3 +123,29 @@ func BenchmarkSyscallWork(b *testing.B) {
 		<-c
 	}
 }
+
+func BenchmarkCreateGoroutines(b *testing.B) {
+	benchmarkCreateGoroutines(b, 1)
+}
+
+func BenchmarkCreateGoroutinesParallel(b *testing.B) {
+	benchmarkCreateGoroutines(b, runtime.GOMAXPROCS(-1))
+}
+
+func benchmarkCreateGoroutines(b *testing.B, procs int) {
+	c := make(chan bool)
+	var f func(n int)
+	f = func(n int) {
+		if n == 0 {
+			c <- true
+			return
+		}
+		go f(n - 1)
+	}
+	for i := 0; i < procs; i++ {
+		go f(b.N / procs)
+	}
+	for i := 0; i < procs; i++ {
+		<-c
+	}
+}