Major enhancements to all multi-core code, bug fixes, performance improvements, etc.

VER### in header.

for multi-prog runs.

Basically works, but needs cleanup and optimization (especially on the
server side).

Also, [srun|slaunch]/multi_prog.c are clearly buggy.

ported as necessary to v1.2.

consistency between the commands.

Convert socket/core/thread info from uint32_t to uint16_t to save some space.

slurmctld yet, but in opt.[ch].

some of the scancel uses lacked "--quiet", which causes a 
failure if the job is already completed. cancel_job also adds 
"wait_for_job DONE" logic.

or env var.