discuss the impact of primary key selection on the performance and scalability of distributed database systems, considering factors such as data partitioning strategies, workload distribution, and the potential for hotspots or contention points within a shared-nothing architecture?