[WIP] ScaFFold Benchmark

[michaelmckinsey1]

Description

• LBANN/ScaFFold experiment implementation. Since the benchmark is private, we set scaffold_path during experiment initialization to know where to look for the clone at runtime.
• Complete the checklist for a relevant section(s) below
• Delete sections below that are not relevant to this PR
• todo:
  • tie pip install to ramble workspace setup pipeline
  • Default benchmark config greater than allocation time, change allocation modifier
  • Add scaling options to experiment
  • Add option for selecting different configurations command-line options

Adding/modifying a benchmark (docs: Adding a Benchmark)

• Add experiments/scaffold/experiment.py, repo/scaffold/application.py

Adding/modifying core functionality, CI, or documentation:

• lib/benchpark/experiment.py: Enable pip package manager variant

git clone https://github.com/LBANN/ScaFFold.git
benchpark experiment init --dest=scaffold scaffold+rocm package_manager=pip scaffold_path="/usr/WS1/mckinsey/ScaFFold/"
benchpark system init --dest=tioga llnl-elcapitan cluster=tioga
benchpark setup scaffold tioga wkp

# Follow ramble instructions
...