Add task_dist (task distribution) to job step information.
Display using the "scontrol show step" command.

The task_dist field was expanded from 16-bits to 32-bits in commit
a4473cf9
However the field size was not expanded in all places and translation
of old format data to the new format was not being done in many
places. This commit, along with commit
45691125
addresses those shortcomings.

The original patch changed RPC formats and contents with respect
to task distribution options, which would have prevented use
with older slurm commands.

dev-hyperthread1-2.patch - Common data structure, option, infrastructure changes

On nodes that support more than one thread per core, each thread can be configured as a separate logical CPU. The srun -m option allows users to specify the method for distributing CPUs across sockets (block, cyclic or fcyclic), but not across cores. The method for distributing CPUs across cores is hardcoded as block. The attached patches implement an extension to the -m option to allow users to request block, cyclic or fcyclic distribution of CPUs across cores. I have also implemented a wildcard character (*) to allow users to request the default value for each distribution type. The new distribution is supported by the task/cgroup plugin only, not task/affinity. The new syntax of the -m option is as follows:

-m, --distribution=
    *|block|cyclic|arbitrary|plane=<options> [:*|block|cyclic|fcyclic[:*|block|cyclic|fcyclic]][,Pack|NoPack]

The main reason for this change is to provide better support for the Xeon Phi architecture, which allows up to four threads per core.
This is dev-hyperthread1-1.patch - Documentation changes

Conflicts:
	src/plugins/switch/cray/util.c

switch/cray: Revert logic added to 14.11.6 that set "PMI_CRAY_NO_SMP_ENV=1"
if CR_PACK_NODES is configured.
bug 1585

If the job step created for external processes (with "PrologFlags=contain")
lacks any accumulated resource use, then don't notify the slurmctld
of completion. This change eliminates an RPC which has no meaningful
data.

Change some job step ID references:
from NO_VAL to SLURM_BATCH_SCRIPT and
from INFINITE to SLURM_EXTERN_CONT

No real change to logic

preempt/job_prio plugin: Implement the concept of Warm-up Time here. Use
the QoS GraceTime as the amount of time to wait before preempting.
Basically, skip preemption if your time is not up.

Conflicts:
	doc/man/man5/slurm.conf.5

Previously was reporting:
Missing node_list for step_id 12727.4294967295

Bug 1548

Conflicts:
	src/common/slurm_protocol_defs.c

Used a value of 3, it should have been a flag, #4