An analytics engineer is building a semantic model over a Fabric lakehouse that contains a 900-million-row sales fact table stored as Delta tables. Business users need near-real-time reporting: data lands in the lakehouse every few minutes, and dashboards must reflect the latest rows without triggering a scheduled import refresh. The team also wants to avoid the cost and maintenance of managing large import partitions, and they require the fastest possible interactive query performance for aggregations. Which storage mode should the engineer choose for the fact table?