DSO+T example has very large memory footprint

[trevorhardy]

The DSO+T cases for one month typically have a memory footprint of 20-40 GB. In some cases, though, we're seeing memory footprints of ~100GB.

[trevorhardy]

HELICS vs FNCS is showing a over 3x memory footprint for GridLAB-D (15GB vs 4.5 GB). In Python processes running at the same time, we see the opposite: FNCS has about a 2x memory footprint as compared to HELICS. All of these are running inside a Docker environment (we don't think this is relevant). We see the results when running the "Rates" memory example and is very repeatable.

[trevorhardy]

As best as we can tell, this does not occur when running on PNNL's HPC where the memory is limited to 64GB. When running these cases on our other machines, they are taking over 100 GB.