Add provenance to inductor IR nodes created after graph.run (#164255)

Summary:
as title

- Some IR nodes are created during `finalize_multi_template_buffers()` in Scheduler. This PR adds provenance (`origin_node` and `origins`) for those nodes.

- Extract `assign_origin_node` function

Differential Revision: D82871244

Pull Request resolved: https://github.com/pytorch/pytorch/pull/164255
Approved by: https://github.com/mlazos

torch/_inductor/scheduler.py

@@ -50,6 +50,7 @@ from .dependencies import Dep, MemoryDep, StarDep, WeakDep
 from .exc import GPUTooOldForTriton, TritonMissing
 from .fx_utils import count_flops_fx
 from .ir import (
+    assign_origin_node,
     get_device_type,
     GraphPartitionSignature,
     MultiOutput,
@@ -3163,12 +3164,16 @@ class Scheduler:
                         node.node.finalize_as_triton_caller(min_node_unfused)
                     continue
 
-                out_tensorbox = min_node_unfused.output_node()
+                with ir.IRNode.current_origins(multi_node.origins):
+                    out_tensorbox = min_node_unfused.output_node()
                 out_storage = out_tensorbox.data  # type: ignore[union-attr]
                 assert isinstance(out_storage, ir.StorageBox)
                 out_buffer = out_storage.data
                 assert isinstance(out_buffer, ir.OperationBuffer)
 
+                if multi_node.origin_node:
+                    assign_origin_node(out_tensorbox, multi_node.origin_node)
+
                 out_buffer.layout = multi_node.layout
                 replace_operation_buffer(multi_node, out_buffer)
                 new_scheduler_node = self.create_scheduler_node(out_buffer)