Use streams for GPU communication

...marks/FlowAroundSphereCodeGen/FlowAroundSphereCodeGen.cpp

apps/benchmarks/NonUniformGridCPU/CMakeLists.txt

apps/benchmarks/NonUniformGridCPU/LdcSetup.h

apps/benchmarks/NonUniformGridGPU/CMakeLists.txt

apps/benchmarks/NonUniformGridGPU/LdcSetup.h

apps/benchmarks/NonUniformGridGPU/NonUniformGridGPU.cpp

...s/NonUniformGridGPU/simulation_setup/benchmark_configs.py

apps/benchmarks/PhaseFieldAllenCahn/benchmark_multiphase.cpp

apps/benchmarks/UniformGridCPU/UniformGridCPU.cpp

...arks/UniformGridCPU/simulation_setup/benchmark_configs.py

apps/benchmarks/UniformGridGPU/CMakeLists.txt

apps/benchmarks/UniformGridGPU/UniformGridGPU.cpp

...arks/UniformGridGPU/simulation_setup/benchmark_configs.py

apps/showcases/PhaseFieldAllenCahn/GPU/multiphase.cpp

apps/tutorials/codegen/03_AdvancedLBMCodegen.cpp

extern/pybind11

src/blockforest/Initialization.cpp

src/geometry/initializer/BoundaryFromDomainBorder.impl.h

src/gpu/communication/NonUniformGPUScheme.h

src/gpu/communication/UniformGPUScheme.h