[WIP] Fix initialisation bug on FeaturePyramidNetwork (pytorch#2954)

* Change children() to modules() to ensure init happens in all blocks.

* Update expected values of all detection models.

* Revert "Update expected values of all detection models."

This reverts commit 050b64a

* Update expecting values.

Author: datumbox

test/expect/ModelTester.test_fasterrcnn_resnet50_fpn_expect.pkl

@@ -86,7 +86,7 @@ def __init__(
             self.layer_blocks.append(layer_block_module)
 
         # initialize parameters now to avoid modifying the initialization of top_blocks
-        for m in self.children():
+        for m in self.modules():
             if isinstance(m, nn.Conv2d):
                 nn.init.kaiming_uniform_(m.weight, a=1)
                 nn.init.constant_(m.bias, 0)