[pytorch/executorch][diff_train] [ez][release blocker fix] Insert linalg_vector_norm into decomp table used for Edge export (#9938)

## Context

Addresses this [release
blocker](https://github.com/orgs/pytorch/projects/99/views/1?pane=issue&itemId=104088363&issue=pytorch%7Cpytorch%7C150207)
issue. Some models cannot export because they use `linalg_vector_norm`
which is not currently an ATen operator.

I initially tried adding the op to the core decomp table, but the decomp
is not passing pytorch correctness tests. Please see
pytorch/pytorch#150241 for more details.

## Changes

Since we currently cannot include the op in PyTorch's decomp table,
instead we can insert the op into the edge decomp table directly.

This PR is a simple change to add `linalg_vector_norm` directly to the
edge decomp table.


Internal:
<< DO NOT EDIT BELOW THIS LINE >>

**GitHub Author**: Sicheng Stephen Jia <ssjia@meta.com> (Meta Employee)
**GitHub Repo**: [pytorch/executorch](https://github.com/pytorch/executorch)
**GitHub Pull Request**: [#9938](#9938)

Initially generated by: https://www.internalfb.com/intern/sandcastle/job/4503601401156690/

This was imported as part of a Diff Train.
Please review this as soon as possible. Since it is a direct copy of a commit on
GitHub, there shouldn't be much to do.

diff-train-source-id: c2e3e17

Differential Revision: [D72729066](https://our.internmc.facebook.com/intern/diff/D72729066/)

[ghstack-poisoned]

Author: kimishpatel

exir/program/test/test_program.py

@@ -725,17 +725,17 @@ def count_nodes(graph_module, target):
         )
 
     def test_edge_dialect_non_core_aten_ops(self):
-        class LinalgNorm(torch.nn.Module):
+        class LinalgRank(torch.nn.Module):
             def __init__(self):
                 super().__init__()
 
             def forward(self, x: torch.Tensor) -> torch.Tensor:
-                return torch.linalg.norm(x)
+                return torch.linalg.matrix_rank(x)
 
         from torch._export.verifier import SpecViolationError
 
-        input = torch.arange(9, dtype=torch.float) - 4
-        ep = torch.export.export(LinalgNorm(), (input,), strict=True)
+        input = torch.ones((9, 9, 9), dtype=torch.float)
+        ep = torch.export.export(LinalgRank(), (input,), strict=True)
 
         # aten::linalg_norm is not a core op, so it should error out
         with self.assertRaises(SpecViolationError):
@@ -748,9 +748,7 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
                 ep,
                 compile_config=EdgeCompileConfig(
                     _check_ir_validity=True,
-                    _core_aten_ops_exception_list=[
-                        torch.ops.aten.linalg_vector_norm.default
-                    ],
+                    _core_aten_ops_exception_list=[torch.ops.aten._linalg_svd.default],
                 ),
             )
         except SpecViolationError:

exir/tracer.py

@@ -631,8 +631,18 @@ def _default_decomposition_table(
         ]
         # pyre-fixme[7]: Expected `Dict[OpOverload, typing.Callable[..., executorch.e...
         return get_decompositions(decomp_opset)
+
+    decomps = default_decompositions()
+    # Add edge specific decompositions
+    additional_decomp_ops = [
+        # TODO: Eventually this op should be added to the core decompo table, and will not
+        # need to be added here.
+        torch.ops.aten.linalg_vector_norm.default,
+    ]
+    additional_decomps = get_decompositions(additional_decomp_ops)
+    decomps.update(additional_decomps)
     # pyre-fixme[7]: Expected `Dict[OpOverload, typing.Callable[..., executorch.exir....
-    return default_decompositions()
+    return decomps
 
 
 def dynamo_trace(