Add parallelism to KeyIndex.aggregate

I looked for serial bottlenecks in sbt project loading and discovered
that KeyIndex.aggregate was relatively easily parallelizable. Before
this time, it took about 1 second to run KeyIndex.aggregate in the akka
project on my computer. After this change, it took 250ms. Given that I
have 4 logical cores, the speedup is roughly linear.

main/src/main/scala/sbt/internal/KeyIndex.scala

@@ -28,10 +28,26 @@ object KeyIndex {
       extra: BuildUtil[_],
       projects: Map[URI, Set[String]],
       configurations: Map[String, Seq[Configuration]]
-  ): ExtendableKeyIndex =
+  ): ExtendableKeyIndex = {
-    (base(projects, configurations) /: known) { (index, key) =>
+    /*
-      index.addAggregated(key, extra)
+     * Used to be:
+     * (base(projects, configurations) /: known) { (index, key) =>
+     *   index.addAggregated(key, extra)
+     * }
+     * This was a significant serial bottleneck during project loading that we can work around by
+     * computing the aggregations in parallel and then bulk adding them to the index.
+     */
+    val toAggregate = known.par.map {
+      case key if validID(key.key.label) =>
+        Aggregation.aggregate(key, ScopeMask(), extra, reverse = true)
+      case _ => Nil
     }
+    (base(projects, configurations) /: toAggregate) {
+      case (index, Nil)  => index
+      case (index, keys) => (index /: keys)(_ add _)
+    }
+  }
+
   private[this] def base(
       projects: Map[URI, Set[String]],
       configurations: Map[String, Seq[Configuration]]