Update on "[ExecuTorch] Add broadcast support for optimized add op"

Summary:
This brings add op to feature parity, wrt, broadcasting, to mul op in
optimized kernels lib

Test Plan:
tests added

Reviewers:

Subscribers:

Tasks:

Tags:

cc larryliu0820 manuelcandales

[ghstack-poisoned]

Author: kimishpatel

kernels/optimized/cpu/op_mul.cpp

@@ -133,8 +133,7 @@ Tensor& opt_mul_out(
     // Reason for using alpha even when used for mul is becasuse
     // handle_broadcast_elementwise is used for add and sub as well
     // and it uses alpha.
-    auto mul_lambda = [](auto x, auto y, auto alpha) {
-      [[maybe_unused]] alpha;
+    auto mul_lambda = [](auto x, auto y, [[maybe_unused]] auto alpha) {
       return x * y;
     };
     static constexpr const char op_name[] = "mul.out";