generic thread pool should always be cached

when using generic thread pool, some elements in the code base rely on the fact that it will always be able to fork it

src/main/java/org/elasticsearch/threadpool/ThreadPool.java

@@ -125,6 +125,9 @@ public class ThreadPool extends AbstractComponent {
             executors.put(executor.getKey(), build(executor.getKey(), groupSettings.get(executor.getKey()), executor.getValue()));
         }
         executors.put(Names.SAME, new ExecutorHolder(MoreExecutors.sameThreadExecutor(), new Info(Names.SAME, "same")));
+        if (!executors.get(Names.GENERIC).info.getType().equals("cached")) {
+            throw new ElasticSearchIllegalArgumentException("generic thread pool must be of type cached");
+        }
         this.executors = ImmutableMap.copyOf(executors);
         this.scheduler = new ScheduledThreadPoolExecutor(1, EsExecutors.daemonThreadFactory(settings, "scheduler"), new EsAbortPolicy());
         this.scheduler.setExecuteExistingDelayedTasksAfterShutdownPolicy(false);

src/test/java/org/elasticsearch/test/TestCluster.java

@@ -232,7 +232,7 @@ public final class TestCluster implements Iterable<Client> {
         }
         if (random.nextBoolean()) {
             // change threadpool types to make sure we don't have components that rely on the type of thread pools
-            for (String name : Arrays.asList(ThreadPool.Names.BULK, ThreadPool.Names.FLUSH, ThreadPool.Names.GENERIC, ThreadPool.Names.GET,
+            for (String name : Arrays.asList(ThreadPool.Names.BULK, ThreadPool.Names.FLUSH, ThreadPool.Names.GET,
                     ThreadPool.Names.INDEX, ThreadPool.Names.MANAGEMENT, ThreadPool.Names.MERGE, ThreadPool.Names.OPTIMIZE,
                     ThreadPool.Names.PERCOLATE, ThreadPool.Names.REFRESH, ThreadPool.Names.SEARCH, ThreadPool.Names.SNAPSHOT,
                     ThreadPool.Names.SUGGEST, ThreadPool.Names.WARMER)) {