[HW][Sim] Add SimulationVariableTypeInterface to let hw.inout hold sim value types

[okekayode]

In this PR we add a marker interface, SimulationVariableTypeInterface, implemented on !sim.assoc_array, !sim.queue, and !sim.dstring. InOutType::verify now accepts either a normal HW value type or this interface, so sv.reg/hw.inout can hold these simulation-only types.

We chose this over widening isHWValueType directly, since it has too many call sites and would let types leak into module ports, hw.array, etc. This keeps the change scope concise ( ports and aggregates still reject these types).

[circt-bot]

Results of circt-tests run for f6cb7d4 compared to results for 6a42084: no change to test results.

[fabianschuiki]

LGTM!

[uenoku]

I’m not sure hw::InOutType is the right abstraction to widen here. This makes types such as !hw.inout<!sim.queue<...>> globally valid, which in turn makes them appear valid for users such as sv.wire, even though queues/strings/associative arrays have variable semantics rather than net semantics. Could we instead model references to SV variables separately, or otherwise keep the existing hw.inout invariant?