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``.

nitransforms/io/lta.py

@@ -162,7 +162,7 @@ def set_type(self, new_type):
     def to_ras(self, moving=None, reference=None):
         """Return a nitransforms internal RAS+ matrix."""
         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."""