Patch from Hongjia Cao

   modules to make use of is_user_any_coord()'s return value.

sbatch initializes ntasks_per_node with 0, while salloc uses NO_VAL.

Both eventually make an allocation based on these parameters. For
consistency, this patch ensures that ntasks_per_node is initialized
with 0 in both programs (there is no 'ntasks_per_node_set', hence
0 was likely meant to indicate 'not set by user').

 allocate.c: In function ‘slurm_allocate_resources_blocking’:
 allocate.c:212: warning: call to function ‘_create_allocation_response_socket’ without a real prototype
 allocate.c:84: note: ‘_create_allocation_response_socket’ was declared here

 filetxt_jobacct_process.c: In function ‘filetxt_jobacct_process_archive’:
 filetxt_jobacct_process.c:1525: warning: call to function ‘slurm_reconfigure’ without a real prototype
 ../../../../slurm/slurm.h:2940: note: ‘slurm_reconfigure’ was declared here

    was set to NULL, caused by the absence of ~/.slurm/sviewrc.

    using the scontrol command.

    in run_job_cnt underflow error message.

    get resumed properly after a slurmctld hot-start.

    and --ntasks-per-node options and the node has fewer CPUs than the count
    specified by --ntasks-per-node.
The bug would prevent test1.37 from completing if the node had fewer than 2 CPUs
    and select/cons_res was used

    some failure modes.

gres records have access to the same CPUs

removed export of LD_LIBRARY_PATH to remove possible security issue with including an empty LD_LIBRARY_PATH which ld.so translates to '.'

    BASIL on Cray computers.

    "node_rank".

    completing job are re-booted. Patch from Hongjia Cao (NUDT).

previously supported).

job step credential to 0.

so that they can be seen using scontrol or squeue.