Improve the precision of several pthread_cond_timedwait() function
calls to be precise to the millisecond rather than rounding to the
neareast second. Applies to most scheduling functions.

Add retry logic related to cpu frequency manament locking.
Some general code improvements also

This should address various race conditions in setting CPU frequency,
especially as relates to gang scheduling (e.g. cancelling or terminating
jobs while near changing the running job).

of the pseudo terminal to the user.

Also don't try to map unset cpu-freq to lowest numeric value (was
looking for NO_VAL rather than zero)

even without it.

appears to be needed for make check

Flesh out logic to aggregate exit state for a job array

together.

line buffered.

Job array dependency based upon state is now dependent upon the state of
the array as a whole (e.g. afterok requires ALL tasks to complete
sucessfully, afternotok is true if ANY tasks does not complete successfully,
and after requires all tasks to at least be started).

Add functions to test the state of a job array, for example
test_job_array_pending() returns true if ANY task of a job array
is pending and test_job_array_completed() returns true if ALL
tasks of a job array are completed.

spaces.

Count is reset running array tasks on reconfig

Recommend use of job arrays and multiple job steps per job.

Cray needs for task_g_post_step() to be called before resetting
the CPU frequency. We also need to reset CPU frequency before
notifying srun of task completion. Logic reorganized to satisfy
this requirement
see bug 1011

sched/backfill - Set expected start time of job submitted to multiple
partitions to the earliest start time on any of the partitions.
Previous logic would set the time to that of the last partition
tested.

This issue was addressed differently and the original 1 second
sleep can be restored for easier performance comparison with
Slurm verstion 14.03.