quickfix: ensure that Dense layers in Sequential are built

The problem seems also to be related to using the Ellipsis in the input
shape.

Fixes #491

Author: vpratz

bayesflow/networks/mlp/mlp.py

@@ -92,6 +92,8 @@ def build(self, input_shape=None):
 
         for layer in self._layers:
             layer.build(input_shape)
+            if not layer.layers[0].built:
+                layer.layers[0].build(input_shape)
             input_shape = layer.compute_output_shape(input_shape)
 
     @classmethod