Conflicts:
	src/plugins/accounting_storage/mysql/as_mysql_acct.c

Set job state to JOB_BOOT_FAIL when boot fails, JOB_NODE_FAIL on
cnode or other hardware failure, otherwise set to JOB_FAILED

In commit ba54c8b4 it was changed
from JOB_FAILED to JOB_NODE_FAIL. JOB_BOOT_FAIL will probably be
more accurate reason

In commit ba54c8b4 the job exit
state was changed to JOB_BOOT_FAIL, which is wrong

bug 403

bug 403

bug 403

bug 403

Make it work on PERCS system.
Add job time limit to avoid indefinite hang
Purge vestigial file from previous run
Exit on fatal condition rather than trying to continue execution

Mostly adding "return" statements after calls to fatal();
bug 403

None of these will happen in produciton
bug 403

Caused by allocating single adapter per node of specific adapter type.

Minor mods to harden Slurm logic
bog 403