Update gcc, MPI and Cuda version in multi gpu tests
Fixes build error in stl_vector.h
that occured when compiling on the Phoenix cluster.
The error message was:
[ 97%] Building CXX object src/gpu/core/CMakeFiles/gpu_core.dir/TurbulenceModels/TurbulenceModelFactory.cpp.o
/cluster/comp/gcc/9.3.0/include/c++/9.3.0/bits/stl_vector.h:385:48: error: ‘_Vp’ was not declared in this scope; did you mean ‘_Tp’?
385 | template<typename _Tp, typename _Alloc = std::allocator<_Tp> >
| ^~~
| _Tp
/cluster/comp/gcc/9.3.0/include/c++/9.3.0/bits/stl_vector.h:385:51: error: template argument 1 is invalid
385 | template<typename _Tp, typename _Alloc = std::allocator<_Tp> >
...