benchpark mirror

[scheibelp]

Closes #566

Say you did:

bin/benchpark experiment init --dest=def-raja-perf raja-perf
bin/benchpark system init --dest=def-ruby llnl-cluster cluster=ruby compiler=gcc
bin/benchpark setup raja-perf-default/ def-ruby/ workspace/
. `pwd`/workspace/setup.sh
ramble --disable-progress-bar --workspace-dir `pwd`/workspace/def-raja-perf/def-ruby/workspace workspace setup

(an example of a typical benchmark workflow that sets up an experiment to run)

Then you can

bin/benchpark mirror create `pwd`/workspace/def-raja-perf/def-ruby/workspace/ test-benchpark-mirror/

move test-benchpark-mirror somewhere, and run

(with test-benchpark-mirror as the CWD)
python3 -m venv mirror-env; . mirror-env/bin/activate
pip install --no-index --find-links=pip-cache pip-cache/*
. `pwd`/setup.sh
ramble --disable-progress-bar --workspace-dir `pwd`/def-raja-perf/def-ruby/workspace/ workspace setup

(note this only provides enough resources in test-benchpark-mirror to run Ramble and Spack, not benchpark itself - doing so would typically not be useful)

[scheibelp]

I'm removing WIP because I can now run the destination-side commands listed above in the description.

[pearce8]

Please add documentation on how to use the mirror capability.