on normal completion.

Now I will really be able to make some progress in squeue. ;-)

Also fixed some problems in the pack methods.

REQUEST_JOB_INFO now uses the job_info_request_msg_t instead of last_update so that in the future I will be able to request information for just one job.
Fixed a lot of bugs in the REQUEST_JOB_STEP_INFO message stuff.

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