Minor tweaks to the crawler test. Added a flush needed under still obscure

high stress situations to prevent a duplicate key violation, in 
PlainDirectoryNodeImpl.  This only makes sense if somehow the second level cache
is getting out of whack.


git-svn-id: https://svn.alfresco.com/repos/alfresco-enterprise/alfresco/BRANCHES/WCM-DEV2/root@3292 c4b6b30b-aa2e-2d43-bbcb-ca4b014f7261

source/java/org/alfresco/repo/avm/PlainDirectoryNodeImpl.java

@@ -190,6 +190,7 @@ class PlainDirectoryNodeImpl extends DirectoryNodeImpl implements PlainDirectory
         if (existing != null)
         {
             existing.setChild(node);
+            sess.flush();  // TODO Should we or shouldn't we?
         }
         else
         {