v1.10.0

What's Changed

• Cross-workspace 4-part naming — models can now read and write to relations in another Fabric workspace via workspace.lakehouse.schema.identifier. Set workspace_name in a model's config() to read across workspaces (#167) or to materialize table and incremental models cross-workspace via CTAS and MERGE INTO (#168). Requires schema-enabled lakehouses; the adapter auto-creates the remote schema on first run.
• Fixed dbt run --full-refresh on incremental Delta models failing with TABLE_OR_VIEW_ALREADY_EXISTS when file_format was not explicitly set — incremental now always drops the existing relation before recreating on full-refresh (#156)
• Fixed mixed-case schema names being lowercased during relation rendering by quoting schema in FabricSparkQuotePolicy (#159)
• Fixed Livy session death (HTTP 404) during query execution to trigger transparent reconnect instead of a hard failure (#159)
• Added missing LivyCursor.fetchmany() and LivySessionConnectionWrapper.fetchone() to complete the DBAPI 2.0 cursor interface, and reset _fetch_index on execute() so re-used cursors no longer return None after the first query (#159)
• Added retry/backoff to create_session() for transient HTTP 404/5xx (Livy not yet available right after lakehouse provisioning) and tolerate transient RequestException/JSONDecodeError in wait_for_session_start() polling (#159)
• Fixed retry warning logs that were silently dropping error details due to an invalid message= kwarg (#159)
• Added REFRESH TABLE before assertions to prevent metastore flake in tests (2772fda)