want instrumentation/"get my state" methods for Propolis entities

Sometimes when debugging live Propolises we'd like to be able to examine the contents of an entity's data structures (e.g. NVMe or Crucible upstairs queues) even when the entity is stuck or otherwise at rest (such that it won't cross any USDT probes or log lines in the course of its continued operation).

One possible way to do this:

• add a trace function to the Entity trait
• each implementation takes locks, examines internal data structures, etc. and logs output and/or fires USDT probes to describe the current state of the system
• Propolis server exposes an API endpoint that visits each entity in turn and calls the tracing callback

The user can then specify which probes to enable, what filtering rules to apply when reading the logs, etc. to get only information from the relevant entities or subsystems.

An alternative approach is to let the user specify exactly which entities should get tracing callbacks. This cuts runtime expenses (no visiting entities you don't care about) in exchange for forcing users to understand how entities were named.

Start by reading the Entity trait, its implementations, and the Propolis server API surface mentioned in the issue. Compare the proposed all-entity callback with targeting specific entities, then define the interface and endpoint behavior. Done means the approach is agreed and current entity state can be inspected through the chosen instrumentation path.