Optimize ByteString.grouped(size) (#25153)

* Add ByteString.grouped() benchmark & tests.

* Implement ByteString.grouped().

* PR comments.

akka-bench-jmh/src/main/scala/akka/util/ByteString_grouped_Benchmark.scala

@@ -0,0 +1,31 @@
+/**
+ * Copyright (C) 2014-2018 Lightbend Inc. <https://www.lightbend.com>
+ */
+
+package akka.util
+
+import java.util.concurrent.TimeUnit
+
+import org.openjdk.jmh.annotations._
+import org.openjdk.jmh.infra.Blackhole
+
+@State(Scope.Benchmark)
+@BenchmarkMode(Array(Mode.AverageTime))
+@OutputTimeUnit(TimeUnit.NANOSECONDS)
+@Warmup(iterations = 5, time = 1, timeUnit = TimeUnit.SECONDS)
+@Measurement(iterations = 10, time = 1, timeUnit = TimeUnit.SECONDS)
+class ByteString_grouped_Benchmark {
+
+  private val bsLarge = ByteString(Array.ofDim[Byte](1000 * 1000))
+
+  /*
+    > akka-bench-jmh/jmh:run -f1 .*ByteString_grouped_Benchmark
+    [info] Benchmark                             Mode  Cnt      Score      Error  Units
+    [info] ByteString_grouped_Benchmark.grouped  avgt   10  59386.328 ± 1466.045  ns/op
+   */
+
+  @Benchmark
+  def grouped(bh: Blackhole): Unit = {
+    bh.consume(bsLarge.grouped(1000).foreach(bh.consume))
+  }
+}