Can you elaborate on the concept of data gravity and its implications for data migration strategies, including considerations such as workload placement, data locality, and latency optimization, particularly in the context of edge computing and decentralized data architectures?