Tests: wait for bulk before closing the test

When closing a test, our test infra removes created indices.

But as the river is still processing documents, the bulk processor contains remaining document which have not been indexed yet.

When the river stops, it fires a `close()` call on the bulk processor.
This call creates again the index. Test framework throw an exception as the mock directory wrapper can not close due to write operations in progress.

Logs:

```
  1> [2014-10-09 11:59:40,965][INFO ][cluster.metadata         ] [node_s0] [_river] deleting index
  1> [2014-10-09 11:59:40,982][INFO ][test.store               ] [node_s1] [_river][5] Shard state before potentially flushing is STARTED
  1> [2014-10-09 11:59:41,084][INFO ][test.store               ] [node_s3] [_river][0] Shard state before potentially flushing is STARTED
  1> [2014-10-09 11:59:41,115][INFO ][test.store               ] [node_s2] [_river][6] Shard state before potentially flushing is STARTED
  1> [2014-10-09 11:59:41,943][INFO ][river.wikipedia          ] [node_s1] [wikipedia][wikipedia] closing wikipedia river
  1> [2014-10-09 11:59:42,693][WARN ][river.wikipedia          ] [node_s1] [wikipedia][wikipedia] river was closing while processing wikipedia page [39]/[Aube (département)
  1>     ]. Operation skipped.
  1> [2014-10-09 11:59:42,768][INFO ][river.wikipedia          ] [WikipediaRiverTest#testWikipediaRiverFrench]: cleaning up after test
  1> [2014-10-09 11:59:42,769][INFO ][cluster.metadata         ] [node_s0] [wikipedia] deleting index
  1> [2014-10-09 11:59:43,587][WARN ][river.wikipedia          ] There was failures while executing bulk
  1> [2014-10-09 11:59:44,971][INFO ][cluster.metadata         ] [node_s0] [wikipedia] creating index, cause [auto(bulk api)], shards [9]/[2], mappings [_default_]
  1> [2014-10-09 11:59:45,805][INFO ][cluster.metadata         ] [node_s0] [wikipedia] update_mapping [page] (dynamic)
  1> [2014-10-09 12:00:00,466][INFO ][test.store               ] [node_s2] [wikipedia][0] MockDirectoryWrapper#close() threw exception
  1> java.lang.RuntimeException: MockDirectoryWrapper: cannot close: there are still open locks: {write.lock=java.lang.RuntimeException: lock "write.lock" was not released}
  1> 	at org.apache.lucene.store.MockDirectoryWrapper.close(MockDirectoryWrapper.java:709)
  1> 	at org.elasticsearch.test.store.MockDirectoryHelper$ElasticsearchMockDirectoryWrapper.close(MockDirectoryHelper.java:140)
  1> 	at org.elasticsearch.test.hamcrest.ElasticsearchAssertions.assertAllFilesClosed(ElasticsearchAssertions.java:667)
  1> 	at org.elasticsearch.test.TestCluster.assertAfterTest(TestCluster.java:76)
  1> 	at org.elasticsearch.test.ElasticsearchIntegrationTest.afterInternal(ElasticsearchIntegrationTest.java:603)
  1> 	at org.elasticsearch.test.ElasticsearchIntegrationTest.after(ElasticsearchIntegrationTest.java:1722)
```

Closes #37.
(cherry picked from commit d2f869c)

src/test/java/org/elasticsearch/river/wikipedia/WikipediaRiverTest.java

@@ -68,18 +68,13 @@ public void createEmptyRiverIndex() {
     public void deleteRiverAndWait() throws InterruptedException {
         logger.info(" --> remove all wikipedia rivers");
         client().admin().indices().prepareDelete("_river").get();
-
+        // We just wait a few to make sure that all bulks has been processed
         awaitBusy(new Predicate<Object>() {
             @Override
             public boolean apply(Object o) {
-                try {
-                    client().admin().indices().prepareGetIndex().addIndices("_river").get();
-                } catch (IndexMissingException e) {
-                    return true;
-                }
                 return false;
             }
-        }, 10, TimeUnit.SECONDS);
+        }, 2, TimeUnit.SECONDS);
     }
 
     @Test