Rename slurm_step_ctx_params_t field from "mem_per_cpu" to "pn_min_memory".
Job step now accepts memory specification in either per-cpu or per-node basis.

Uninitialized variables resulted in error of
"cons_res: sync loop not progressing, holding job #"

calls only took ~1500 usec to complete call.  Since this is out of band
this shouldn't be that big of a deal.

The most notable problem case is on a cray where a job step
specifically requests one or more node that are not the first
nodes in the job allocation

If node_name2bitmap() is called with best_effort=false, then do
not attempt to match names with NodeHostName.

Without this change, a partition that contains a NodeHostName rather
that NodeName would be configured with the first one found. On a
front-end system, this would result in the partition's node_bitmap
being out of sync with the actual node positions.

To reproduce the problem, configure with --enable-multiple-slurmd
Then in slurm.conf, define something like this:
NodeName=foo[1-8] NodeHostName=bar ...
PartitionName=debug Nodes=bar,foo[1-8] ...

a limit issue.

bug 306 fix