[Executorch][llm] Compile custom op with -O2 (#6587)

Pull Request resolved: #6518

This enable some optimizations for inlining vectorized lib functions
ghstack-source-id: 251021622
@exported-using-ghexport

Differential Revision: [D64910576](https://our.internmc.facebook.com/intern/diff/D64910576/)

Co-authored-by: Kimish Patel <[email protected]>

Author: pytorchbot

extension/llm/custom_ops/targets.bzl

@@ -1,4 +1,9 @@
 load("@fbsource//xplat/executorch/build:runtime_wrapper.bzl", "runtime")
+load(
+    "@fbsource//xplat/executorch/kernels/portable:op_registration_util.bzl",
+    "get_compiler_optimization_flags",
+)
+
 
 def define_common_targets():
     """Defines targets that should be shared between fbcode and xplat.
@@ -34,7 +39,7 @@ def define_common_targets():
                 "//executorch/kernels/portable/cpu/util:reduce_util",
                 "//executorch/extension/llm/custom_ops/spinquant:fast_hadamard_transform",
             ],
-            compiler_flags = ["-Wno-missing-prototypes", "-Wno-global-constructors"],
+            compiler_flags = ["-Wno-missing-prototypes", "-Wno-global-constructors"] + get_compiler_optimization_flags(),
             visibility = [
                 "//executorch/...",
                 "//executorch/extension/llm/custom_ops/...",