Update on "[Executorch] Add broadcasting support to optimized op_sub"

Summary:
This diff builds on top of previous one to add support for limited
handling of broadcasting for sub

Test Plan:
tests added

Reviewers:

Subscribers:

Tasks:

Tags:

cc larryliu0820 manuelcandales

Differential Revision: [D69491818](https://our.internmc.facebook.com/intern/diff/D69491818)

[ghstack-poisoned]

Author: kimishpatel

kernels/optimized/cpu/binary_ops.h

@@ -199,9 +199,7 @@ Tensor& handle_last_dim_broadcast_elementwise(
     const Tensor& a,
     const Tensor& b,
     Tensor& out,
-    const ElementwiseOptimizedPath selected_optimized_path,
-    const executorch::aten::optional<Scalar>& alpha = {}) {
-  ScalarType out_type = out.scalar_type();
+    const ElementwiseOptimizedPath selected_optimized_path) {
   const Tensor* lhs;
   const Tensor* rhs;
   if (selected_optimized_path ==