Update on "[Executorch] Renable operator optimization flags"

Previous attempt at this resulted in revert due to app size increase. Much of
this was due to op_div exploding. Two diffs underneath solve this issue

Differential Revision: [D65606666](https://our.internmc.facebook.com/intern/diff/D65606666/)

[ghstack-poisoned]

Author: kimishpatel

shim/xplat/executorch/kernels/portable/op_registration_util.bzl

@@ -1279,9 +1279,16 @@ CUSTOM_OPS = (
     ),
 )
 
+# Lists which ops are refactored into separate impl files that can be reused elsewhere.
+# Not the best way to track this since we should not have list files like this at all
+# but maybe we can figure out how to refactor selective build.
+ATEN_OP_IMPLS = ["op_div"]
+
 def portable_source_list():
     """All the source file names from //executorch/kernels/portable/cpu/"""
-    return [op["name"] + ".cpp" for op in ATEN_OPS + CUSTOM_OPS]
+    sources = [op["name"] + ".cpp" for op in ATEN_OPS + CUSTOM_OPS]
+    sources = sources + [op + "_impl.cpp" for op in ATEN_OP_IMPLS]
+    return sources
 
 def portable_header_list():
     """All the header file names from //executorch/kernels/portable/cpu/"""