Design and implement idealized CLI for user-experience

[vprusso]

We presently dispatch specific benchmarking jobs like this (for instance):

python metriq_gym/dispatch_qv.py --shots 1 --backend ibm_nazca --trials 1 --n 5 --provider ibmq

Likewise, we poll specific benchmarking jobs like this (for instance):

python metriq_gym/poll_qv.py  

However, we would ideally like to provide "dispatch" and "poll" as agnostic options that can be supplied as arguments via the CLI and not necessarily tie them to specific benchmarks. So, for example, one might instead do this for dispatching:

metriq-gym -dispatch --benchmark quantum_volume --provider ibmq --backend ibm_nazca --num_qubits 5 --shots 1

This would generate a specific metriq-gym job ID. This ID could then be used to retrieve the job status and (if finished) the results of the job.

metriq-gym -retrieve --id 1234

[cosenal]

Fixed by #81