implement dpnp.bitwise_count

[vtavana]

In this PR, dpnp.bitwise_count is implemented.

• Have you provided a meaningful PR description?
• Have you added a test, reproducer or referred to issue with a reproducer?
• Have you tested your changes locally for CPU and GPU devices?
• Have you made sure that new changes do not introduce compiler warnings?
• Have you checked performance impact of proposed changes?
• If this PR is a work in progress, are you filing the PR as a draft?

[github-actions]

View rendered docs @ https://intelpython.github.io/dpnp/pull/2308/index.html

[github-actions]

Array API standard conformance tests for dpnp=0.17.0dev6=py312he4f9c94_22 ran successfully.
Passed: 991
Failed: 1
Skipped: 22

[antonwolfy]

Do we need to have a vector implementation, since sycl::popcount support that?
And we have options here: it can be supported either only for int8_t where no casting is needed or for all integer types through explicit vector casting, like dpctl does:

    template <int vec_sz>
    sycl::vec<resT, vec_sz> operator()(const sycl::vec<argT, vec_sz> &x) const
    {
        if constexpr (std::is_unsigned_v<argT>) {
            auto const &res_vec = sycl::popcount(x);

            using deducedT = typename std::remove_cv_t<
                    std::remove_reference_t<decltype(res_vec)>>::element_type;

            return vec_cast<std::uint8_t, deducedT, vec_sz>(res_vec);
        }
        else {
            auto const &res_vec = sycl::popcount(sycl::abs(x));

            using deducedT = typename std::remove_cv_t<
                    std::remove_reference_t<decltype(res_vec)>>::element_type;

            return vec_cast<std::uint8_t, deducedT, vec_sz>(res_vec);
        }
    }

The question only if any of that will bring performance benefits.

[ndgrigorian]

I would be curious to know as well.

In the dpctl PR which started work on adding vector overloads for unary functions (IntelPython/dpctl#1223), little benefit was found, subgroup store/load seemed to make much more of a difference.

[coveralls]

coverage: 71.704% (-0.03%) from 71.737%
when pulling 371319a on impl-bitwise_count
into 0d5ffad on master.