Remove call to scheduling loop on failed job submit RPC (redundant).

added %m to a lot of debug messages

Added support for combined allocate resources and run job step API.
Changed allocate_tst.c to perform allocate_and_run RPC on first request.

Added support for combined allocate resources and run job step API.

moved things around and cleaned up things

Begining of cleanup of slurmd

should be MESSAGE_TASK_EXIT

moved launch_tasks_response_msg to be sent before
func call that blocks until tasks exit

slurmd now properly exits io threads on task exit and sends a task exit message back to srun

slurmd should also now accept job_credential_revoke messages, althought it doesn't currently process them

   of directly to a stream.

 o various small bugfixes.

Move getnodename to common/hostlist, function is equivalent to
     gethostname, but strips off everything after first "."
slurmd to use getnodename now

Move getnodename to common/hostlist, function is equivalent to
     gethostname, but strips off everything after first "."

Translate "localhost" in node configuration inforo and partition
node list to actual node name.
Reduce size on allocate test to something that will run on PC.