Refactor layout constraint selection logic (#148104)

This PR:

- cleans up some existing comments that don't make sense anymore
- hooks up the "custom_op_default_layout_constraint" back (that seems to
have broken)
- cleans up the "lazy registration path" which seems to never get hit
anymore
- adds dislike_padding to nodes that require exact strides

Test Plan:
- tests + CI

disable padding

Pull Request resolved: https://github.com/pytorch/pytorch/pull/148104
Approved by: https://github.com/shunting314, https://github.com/eellison

test/test_custom_ops.py

@@ -3849,7 +3849,7 @@ Please use `add.register_fake` to add an fake impl.""",
             # If multiple tags are provided, then we use the most constrained tag.
             ({flexible_layout, needs_fixed_stride_order}, needs_fixed_stride_order),
         ]
-        from torch._inductor.lowering import get_layout_constraint_tag
+        from torch._library.utils import get_layout_constraint_tag
 
         for tags, expected in tests:
             with torch.library._scoped_library("mylib", "FRAGMENT") as m: