Add a new API in variorum_topology.h for coprocessors (e.g. GPUs)

[tpatki]

Some helpful notes from the Flux team (thank you Tom and Mark!).

• How Flux handles coprocessors: https://github.com/flux-framework/flux-core/blob/596b073199aec51aae6c89118644d8dd57277975/src/common/librlist/rhwloc.c#L268-L303

• Some helpful notes specific to Lassen here: https://flux-framework.readthedocs.io/en/latest/tutorials/lab/coral.html?h=lassen#scheduling-gpus

[tpatki]

At present, for v0.8, we are getting past this issue by exposing num_gpus_per_socket in the JSON object returned from the variorum_get_power_json API.