[SYCL][COMPAT] Util header updated to latest version

[Alcpz]

The util header is divided in utility functions and math functions to help with machine translated code.

[zhiweij1]

Hi @Alcpz , we recently found an issue in these min/max overloads.
In this implementation, we use std::int??_t/uint??_t, but these types are not fundamental types.
If user pass 2 longlong type variables to dpct::max, and if int64_t is typedef of long, then compiler does not know which overload should be chosen.
And when we want to add a new longlong overload, it is difficult to know if there is redefinition since which type is chosen to define int64_t is implementation defined. (gcc treats int64_t and long are same, but msvc treats int64_t and longlong are same).
So I suggest using fundamental types (int/long/longlong/...) here instead of fixed width integer types.

[zhiweij1]

Or is something like below better?

template <typename T1, typename T2>
std::enable_if_t<std::is_integral_v<T1> && std::is_integral_v<T2>,
                 std::common_type_t<T1, T2>>
min(T1 a, T2 b) {
  return sycl::min<std::common_type_t<T1, T2>>(a, b);
}
template <typename T1, typename T2>
std::enable_if_t<std::is_floating_point_v<T1> && std::is_floating_point_v<T2>,
                 std::common_type_t<T1, T2>>
min(T1 a, T2 b) {
  return sycl::fmin<std::common_type_t<T1, T2>>(a, b);
}
template <typename T1, typename T2>
std::enable_if_t<std::is_integral_v<T1> && std::is_integral_v<T2>,
                 std::common_type_t<T1, T2>>
max(T1 a, T2 b) {
  return sycl::max<std::common_type_t<T1, T2>>(a, b);
}
template <typename T1, typename T2>
std::enable_if_t<std::is_floating_point_v<T1> && std::is_floating_point_v<T2>,
                 std::common_type_t<T1, T2>>
max(T1 a, T2 b) {
  return sycl::fmax<std::common_type_t<T1, T2>>(a, b);
}

[Alcpz]

Thanks for pointing this @zhiweij1. I will add the changes you propose once I find time to get back to these PRs.

[Alcpz]

Since this PR is closed, I will ping you in the appropriate PR that contains these changes

[Alcpz]

This PR grew too much in scope. In order to faciliate review, it is being splitted beginning with #12957