fix: set large_dtypes=False as the default behavior on merge

Automatic upcasting of the data type is surprising and unexpected for
users. IMHO we should not be casting data types unless required

python/deltalake/table.py

@@ -825,7 +825,7 @@ def merge(
         target_alias: Optional[str] = None,
         error_on_type_mismatch: bool = True,
         writer_properties: Optional[WriterProperties] = None,
-        large_dtypes: bool = True,
+        large_dtypes: bool = False,
         custom_metadata: Optional[Dict[str, str]] = None,
     ) -> "TableMerger":
         """Pass the source data which you want to merge on the target delta table, providing a

python/tests/test_merge.py

@@ -970,17 +970,18 @@ def test_struct_casting(tmp_path: pathlib.Path):
     assert not df.empty
 
     schema = pa.Table.from_pandas(df=df).schema
-
     dt = DeltaTable.create(tmp_path, schema, name="test")
     metadata = dt.metadata()
     assert metadata.name == "test"
 
     result = (
         dt.merge(
-            source=df_merge, predicate="t.id = s.id", source_alias="s", target_alias="t"
+            source=df_merge,
+            predicate="t.id = s.id",
+            source_alias="s",
+            target_alias="t",
         )
         .when_matched_update_all()
-        .when_not_matched_insert_all()
         .execute()
     )
     assert result is not None