[doc] Update following the changes in stdlib.

doc/mnist.rst

@@ -86,13 +86,13 @@ the local robustness property:
 
 .. code-block:: whyml
 
-  predicate robust_at (m: model) (d: datum) (eps: t) =
+  predicate robust_at (m: model) (d: record) (eps: t) =
     forall x': features.
       let (x, _) = d in
       linfty_distance x x' eps ->
       predict m x = predict m x'
 
-Note that a ``datum`` is a dataset element given as a pair of *features* and
+Note that a ``record`` is a dataset element given as a pair of *features* and
 (classification) *label*. Morever, ``linfty_distance`` is a predicate that
 describes how two arrays of floating-point values (*i.e.* ``features``) are
 considered close up-to a pertubation ``eps``, while ``predict`` is a function