fixed some error codes in Ndb so that 4009-cluster failure is returned when cluster is not up

fixed memory leak in object map
fixed uninitialized value blobs_pending


ndb/include/mgmapi/mgmapi.h:
  added ndb_mgm_destroy_configuration call
ndb/src/mgmapi/mgmapi.cpp:
  added ndb_mgm_destroy_configuration call
ndb/src/ndbapi/ObjectMap.hpp:
  fixed memory leak
ndb/src/ndbapi/ndb_cluster_connection.cpp:
  added ndb_mgm_destroy_configuration call
sql/ha_ndbcluster.cc:
  fixed uninitialized value blobs_pending

some other small fixes


ndb/include/mgmcommon/ConfigRetriever.hpp:
  added options to do_connect to contol how connects failures should be treated
ndb/include/mgmcommon/NdbConfig.h:
  method to retrieve datadir path (to user for chdir)
ndb/include/ndbapi/ndb_cluster_connection.hpp:
  Added connect thread
ndb/src/common/mgmcommon/ConfigRetriever.cpp:
  added options to do_connect to contol how connects failures should be treated
ndb/src/common/mgmcommon/NdbConfig.c:
  method to retrieve datadir path (to user for chdir)
ndb/src/kernel/main.cpp:
  ndbd to do chdir
ndb/src/kernel/vm/WatchDog.cpp:
  added my_thread_init for debug
ndb/src/ndbapi/ClusterMgr.cpp:
  added my_thread_init for debug
ndb/src/ndbapi/TransporterFacade.cpp:
  removed call to atexit
ndb/src/ndbapi/ndb_cluster_connection.cpp:
  added connect thread

added class Ndb_cluster_connection to prepare for making order of starting mysqld and ndbd irrelevant


ndb/src/mgmsrv/Makefile.am:
  changed so that datadir is "." by default so that behaviour is the same as before