bench: remove manual GC from garbage benchmark
Initially the garbage benchmark was flaky because
some runs trigger 3 GCs while others trigger 4 GCs.
This caused 33% flakes.
Then I've added the manual GC triggering that emulates
GOGC=100 but more stable. This fixed the flakiness.
Then we've changed runtime.GC so that it executes 2 GCs.
This increased GC time by 50%, and numbers between releases
become uncomparable.
I don't have a better idea than to remove the manual GC triggerring.

LGTM=adg
R=adg
CC=golang-codereviews, rsc
https://golang.org/cl/88040051
diff --git a/driver/driver_go10.go b/driver/driver_go10.go
index 8d2a3f4..2874b98 100644
--- a/driver/driver_go10.go
+++ b/driver/driver_go10.go
@@ -13,7 +13,3 @@
 // New mem stats added in Go1.2
 func collectGo12MemStats(res *Result, mstats0, mstats1 *runtime.MemStats) {
 }
-
-func SetGCPercent(x int) bool {
-	return false
-}
diff --git a/driver/driver_go12.go b/driver/driver_go12.go
index 87b3dad..8dfed01 100644
--- a/driver/driver_go12.go
+++ b/driver/driver_go12.go
@@ -8,7 +8,6 @@
 
 import (
 	"runtime"
-	"runtime/debug"
 )
 
 // New mem stats added in Go1.2
@@ -16,9 +15,3 @@
 	res.Metrics["sys-gc"] = mstats1.GCSys
 	res.Metrics["sys-other"] = mstats1.OtherSys + mstats1.MSpanSys + mstats1.MCacheSys + mstats1.BuckHashSys
 }
-
-func SetGCPercent(x int) bool {
-	debug.SetGCPercent(x)
-	runtime.GC()
-	return true
-}
diff --git a/garbage/garbage.go b/garbage/garbage.go
index 31bccb3..816adb5 100644
--- a/garbage/garbage.go
+++ b/garbage/garbage.go
@@ -26,8 +26,7 @@
 type ParsedPackage map[string]*ast.Package
 
 var (
-	parsed   []ParsedPackage
-	manualGC bool
+	parsed []ParsedPackage
 )
 
 func benchmark() driver.Result {
@@ -42,11 +41,6 @@
 			}
 		}
 		fmt.Printf("consumption=%vKB npkg=%d\n", mem>>10, npkg)
-		// Disable GC, we will trigger GC manually at predictable points.
-		// This helps to avoid a situation when one run triggers 3 GCs,
-		// and then a subsequent run triggers 4 GCs, which causes
-		// significant fluctuations in results.
-		manualGC = driver.SetGCPercent(10000)
 	}
 	return driver.Benchmark(benchmarkN)
 }
@@ -72,10 +66,6 @@
 				// the other part represents "old" generation.
 				parsed[pos%(len(parsed)/2)] = p
 				pos++
-				// This condition is roughy equal to GCGC=100, but is more stable.
-				if manualGC && (pos%len(parsed)) == 0 {
-					runtime.GC()
-				}
 				mu.Unlock()
 				<-gate
 			}