Update on "[kernel] Add template based unboxing"

Adding a new feature to allow users to bypass codegen and register their
kernels directly. This is very useful for custom kernels for custom ops.

Example usage:
```
Tensor& my_op(RuntimeContext& ctx, const Tensor& self, const Tensor& other, Tensor& out) {
    // ...
    return out;
}

Kernel my_kernel = Kernel.make_boxed_kernel("my_ns::my_op",EXECUTORCH_FN(my_op));

register_kernels({my_kernel});
```

Differential Revision: [D51553099](https://our.internmc.facebook.com/intern/diff/D51553099)

[ghstack-poisoned]

Author: larryliu0820

extension/kernel_util/meta_programming.h

@@ -6,7 +6,6 @@
  * LICENSE file in the root directory of this source tree.
  */
 
-
 #pragma once
 #if __cplusplus < 201703L
 #error "This header requires C++17"

extension/kernel_util/type_list.h

@@ -15,6 +15,7 @@
 #error "This header requires C++17"
 #endif
 
+#include <algorithm>
 #include <cstddef>
 #include <type_traits>
 #include <utility>
@@ -131,9 +132,9 @@ struct drop_if_nonempty final {
       "In typelist::drop<T, num>, the T argument must be typelist<...>.");
   using type = typename take_elements<
       TypeList,
-      min(num, size<TypeList>::value),
+      std::min(num, size<TypeList>::value),
       std::make_index_sequence<
-          size<TypeList>::value - min(num, size<TypeList>::value)>>::type;
+          size<TypeList>::value - std::min(num, size<TypeList>::value)>>::type;
 };
 template <class TypeList, size_t num>
 using drop_if_nonempty_t = typename drop_if_nonempty<TypeList, num>::type;