Cbench - Scalable Cluster Benchmarking Code

have to use cbench_echo for all echos

dump out some basic info about modules loaded,
libary paths, and such as it exists inside the
job

make sure dplot is installed if just 'make nodehwtest' was
used

--nodediag option needed updating to deal with jobs that have
a RUNNING status (only happens with Slurm right now)

don't even call open_and_slurp() on output files for jobs in
the RUNNING state

tweaks to --nodediag output

another variant for nehalem

Add a stream job that uses ~11 GB of memory

Add Cbench support for sPPM

Add sPPM to Cbench openapps.

Problem size is statically built into the sppm executable, so any modifications to problem size require a rebuild. Currently, only one binary is built and it has a problem size that uses ~1.2-1.4 GB of memory per thread.

Runtime is set in the inputdeck.cbench file and is currently set to 100 timesteps. Runs of this length and size typically take 1-3+ hours to complete.

bugfix

add the miranda_io test to the IO testset