aim is to:
  a) if set_connect_timeout called, timeout connect attempt (for retry on
  next call) after timeout period
  b) preserve existing blocking behaviour otherwise (for, e.g. mgmapi)
  
  Related to customer issue with long time deleting ndb_cluster_connection
  object. believe we're hanging on the connect(2) call until timeout (when
  we then realise we should exit the thread).


ndb/include/mgmapi/mgmapi.h:
  add ndb_mgm_set_connect_timeout
ndb/include/util/SocketClient.hpp:
  add timeout (seconds) for max time to wait for connection
ndb/src/common/transporter/Transporter.cpp:
  set limit on amount of time we'll wait for tcp connect
ndb/src/common/util/SocketClient.cpp:
  only try to connect for a maximum of timeout time
ndb/src/mgmapi/mgmapi.cpp:
  add ndb_mgm_set_connect_timeout