`benchpark setup`: easier referencing of dynamically generated systems

[scheibelp]

benchpark setup, when using a dynamically generated system (from benchpark system init), will generate a ramble workspace in part using the name/hash of the system

This can make it awkward to reference, as is demonstrated in our dry run CI checks:

./bin/benchpark system init --dest=tioga-system tioga rocm=551 compiler=cce ~gtl
...
./bin/benchpark setup ./saxpy-rocm ./tioga-system workspace/
...
ramble --workspace-dir workspace/saxpy-rocm/Tioga-975af3c/workspace 
^
|
(this has to refer to "Tioga-975af3c")

This adds some convenience functionality:

1. New command benchpark system id <a dir generated by system-init>: which will print the directory component related to the system (e.g. "Tioga-975af3c" in the above example)
2. benchpark setup now generates a symlink that follows the same pattern as for an experiment, using the basename of the system config dir. In terms of the above example, that would be workspace/saxpy-rocm/tioga-system -> workspace/saxpy-rocm/Tioga-975af3c

Notes:

• Keeping the hash-based link would help users that want to dynamically generate multiple workspace from a set of (experiment/system) tuples (the symlink would have to be removed for that to work well though)
• Until then (if ever) the symlink based on the simple name means a user can anticipate the name of the Ramble workspace generated by benchpark setup
• Either way, keeping the hash-name means that any existing scripts referring to system config dirs will continue to work
  • ... if that is important
  • if not, then this could just switch over to the simple name (i.e. the symlink becomes the actual path, and the hash is omitted)

[pearce8]

Resolves #443.