fix: seems like LTA affines are defined inversed

In other words, the ``src volume`` is the moving image and the
``dest volume`` is the reference image. However, the matrix written in
the file seems to map coordinates from ``src`` to ``dest`` (i.e., it is
the right one to move the data from ``dest`` to ``src``.

Author: oesteban

nitransforms/io/lta.py

@@ -160,9 +160,38 @@ def set_type(self, new_type):
         )
 
     def to_ras(self, moving=None, reference=None):
-        """Return a nitransforms internal RAS+ matrix."""
+        """
+        Return a nitransforms' internal RAS+ array.
+
+        Seemingly, the matrix of an LTA is defined such that it
+        maps coordinates from the ``dest volume`` to the ``src volume``.
+        Therefore, without inversion, the LTA matrix is appropiate
+        to move the information from ``src volume`` into the
+        ``dest volume``'s grid.
+
+        .. important ::
+
+            The ``moving`` and ``reference`` parameters are dismissed
+            because ``VOX2VOX`` LTAs are converted to ``RAS2RAS`` type
+            before returning the RAS+ matrix, using the ``dest`` and
+            ``src`` contained in the LTA. Both arguments are kept for
+            API compatibility.
+
+        Parameters
+        ----------
+        moving : dismissed
+            The spatial reference of moving images.
+        reference : dismissed
+            The spatial reference of moving images.
+
+        Returns
+        -------
+        matrix : :obj:`numpy.ndarray`
+            The RAS+ affine matrix corresponding to the LTA.
+
+        """
         self.set_type(1)
-        return self.structarr['m_L']
+        return np.linalg.inv(self.structarr['m_L'])
 
     def to_string(self):
         """Convert this transform to text."""