Add cuda-compatible ippl::Tuple with test

[manuel5975p]

Add __device__-compliant ippl::Tuple type, inspired by std::tuple

Currently only in one file Types/Tuple.h, since most functions are trivial one-liners.

Some examples:

{
    ippl::Tuple<int, std::unique_ptr<int>> tuple1(17, std::make_unique<int>(5));
    ippl::Tuple<int, std::unique_ptr<int>> tuple2{5, nullptr};
    //tuple2 = tuple1 is ill-formed
    tuple2 = std::move(tuple1); //Moves correctly
    
    ippl::Tuple<int, double> values1{5, 10.0};
    ippl::Tuple<int, double> values2(0.5f, 1.0f); //Both work, 0.5f is casted to 0 (type int).

    std::cout << ippl::get<0>(values1 + values2); // 5
    std::cout << ippl::get<1>(values1 - values2); // 9.0
}

Arithmetic operators are only enable for tuples in which all types satisfy is_arithmetic

Tests work for Serial, OpenMP and Cuda, including a parallel_for and a parallel_reduce loop.

A usecase for this is Boundary Conditions for FDTD (and also general BCs) since the axis along which the wave vanishes should be a template parameter.

[Arc676]

Could use some additional empty lines to space out some things but otherwise seems good.

[Arc676]

I would split this macro over several lines for readability.

However, this program also feels more like a unit test. Instead of building your own assert, you could just use the ones already in GTest.

[manuel5975p]

Ah yes, that was a bit of a hack, also I just went off tests/vector

[Arc676]

@aaadelmann What do you think? I would argue that the nature of this program makes it more suited to be included as a unit test. I had previously argued in favor of trimming most of the test/ programs anyway in favor of converting them to more helpful, illustrative example programs and restricting actual internal testing to either the unit tests or programs that are not tracked with the repository, or just minimal testbeds.

[aaadelmann]

@manuel5975p please open an issue for the unit test! @Arc676 please approve