Table: Fix ensure_copy for sparse matrices

`ensure_copy` is checking whether an array is a view through the `.base` argument which doesn't exist on `scipy.sparse` matrices. Since `scipy.sparse` don't work with views (e.g. indexing returns a copy of the matrix) the check should only be performed for dense matrices.

Orange/data/table.py

@@ -912,15 +912,16 @@ def is_copy(self):
 
     def ensure_copy(self):
         """
-        Ensure that the table owns its data; copy arrays when necessary
+        Ensure that the table owns its data; copy arrays when necessary.
+        The check is skipped for sparse matrices since they don't have views like numpy arrays.
         """
-        if self.X.base is not None:
+        if not sp.issparse(self.X) and self.X.base is not None:
             self.X = self.X.copy()
-        if self._Y.base is not None:
+        if not sp.issparse(self._Y) and self._Y.base is not None:
             self._Y = self._Y.copy()
-        if self.metas.base is not None:
+        if not sp.issparse(self.metas) and self.metas.base is not None:
             self.metas = self.metas.copy()
-        if self.W.base is not None:
+        if not sp.issparse(self.W) and self.W.base is not None:
             self.W = self.W.copy()
 
     def copy(self):