Use ParallelFor in Value class

#pragma once

#include <TNL/Algorithms/ParallelFor.h>

#include <TNL/Containers/Array.h>

#include <memory>

    static TNL::Containers::Array<Value<T>, Device> fill(const Array<T, Device>& values) {

        TNL::Containers::Array<Value<T>, Device> arr;

        arr.setSize(values.getSize());

        for (int i = 0; i < values.getSize(); i++)

            arr[i] = Value(values[i]);

        //for (int i = 0; i < values.getSize(); i++)

        auto aView = arr.getView();

        auto vView = values.getConstView();

        auto _fill = [aView, vView] __cuda_callable__ (int i) mutable {

            aView[i] = Value(vView[i]);

        };

        TNL::Algorithms::ParallelFor<Device>::exec(0, values.getSize(), _fill);

        return std::move(arr);

    }