Is your feature request related to a problem? Please describe.

I would like to be able to specify a "default" resource per GPU device available to my process. Today, RMM only has a concept of a single default resource.

Describe the solution you'd like

Create a mechanism to specify/manage a set of default resources, one per device.

Additional context
Thrust has a concept of a per_device_resource by managing a std::map<device_id, memory_resource>.