is exec'ed.  This implementation is not ideal, because for switch/federation
the slurmd_step process needs to read in and parse the federation.conf file
every time the slurmd_step is exec'ed.

Added "switch_g_slurmctld_init", "switch_g_slurmd_init", and
"switch_g_slurmd_step_init" calls to the switch API.

NOTE:  The "switch_g_slurmctld_init" function is not yet called anywhere in
the slurmctld, but it also is not yet needed by any switch plugin.

  slurm_container_create() and slurm_container_add().  One of those
  functions MUST set job->cont_id.

o Add working proctrack/rms plugin.

    (Chris Holmes, HP).

to start with instead of always using 1, which is prone to 
collisions.

signed to/from unsigned variables, variable precision, etc.).

QSW library logic).

node-by-node basis rather than tracking use on a global basis.

    rather than just using a cyclic counter and hoping to avoid collisions.

    of communication protocol over IBM Federation switch.
 -- Add new slurm.conf parameter ProctrackType (process tracking) and
    associated plugin in the slurmd module.

tests were change appropriately. This lets one fake a Linux cluster
without all of the BGL-specific logic.

Not critical, but a loose end in found in tracking memory leaks.

 uninitialized variables (thanks to valgrind).

  switch.

within the RMS context range. This avoid Elan intallation
errors with both lowcontext and highcontext are not with this
range. (gnats:482)

No change for non-AIX systems.

stack size to an appropriate size. System dies on AIX without this due
to too small of a stack.

   libelanhosts library.
 o Test for libelanhosts at configure time.