Wrap gRPC StatusRuntimeException across all client and worker gRPC calls

[bachuv]

Summary

Most gRPC calls in the Java SDK let raw io.grpc.StatusRuntimeException escape to callers. The .NET SDK wraps RpcException (at minimum StatusCode.Cancelled → OperationCanceledException) for all entity client methods. The Java SDK should provide consistent domain-level exception wrapping so callers don't need to depend on gRPC types and so we can have .NET feature parity.

Current Behavior

The majority of gRPC methods across DurableTaskGrpcClient, GrpcDurableEntityClient, and DurableTaskGrpcWorker let StatusRuntimeException propagate unwrapped:

DurableTaskGrpcClient — no wrapping:

• suspendInstance()
• resumeInstance()
• terminate()
• raiseEvent()
• createInstanceWithRaise()
• getInstanceMetadata()

DurableTaskGrpcClient — wraps specific codes only:

• waitForInstanceStart() — DEADLINE_EXCEEDED → TimeoutException
• waitForInstanceCompletion() — DEADLINE_EXCEEDED → TimeoutException
• purgeInstances() — DEADLINE_EXCEEDED → TimeoutException
• rewindInstance() — NOT_FOUND → IllegalArgumentException, FAILED_PRECONDITION → IllegalStateException

GrpcDurableEntityClient — no wrapping:

• signalEntity()
• getEntityMetadata()
• queryEntities()
• cleanEntityStorage()

DurableTaskGrpcWorker — no wrapping:

• completeActivityTask()
• completeOrchestratorTask()
• completeEntityTask()

Expected Behavior

gRPC transport exceptions should be translated into SDK-level exceptions. At minimum:

• CANCELLED → CancellationException (closest Java equivalent to .NET's OperationCanceledException)
• NOT_FOUND → IllegalArgumentException or a dedicated not-found exception
• DEADLINE_EXCEEDED → TimeoutException (already done for some methods)

.NET SDK Reference

The .NET GrpcDurableEntityClient wraps RpcException for every entity method:

catch (RpcException e) when (e.StatusCode == StatusCode.Cancelled)
{
    throw new OperationCanceledException(
        $"The {nameof(this.SignalEntityAsync)} operation was canceled.", e, cancellation);
}