re-add cuda docker container

[chapman39]

now that serac is being installed with nvcc everywhere, the cuda build has quite a few issues that should be fixed before re-adding the docker container.

"trial_spaces" is undefined in device code

https://github.com/LLNL/serac/blob/develop/src/serac/numerics/functional/shape_aware_functional.hpp#L421

when compiling solid_multi_material.cpp (and some other tests) with nvcc, i get the following error:

/home/radiuss/serac_repo/src/serac/infrastructure/../../serac/numerics/functional/shape_aware_functional.hpp(421): error: identifier "serac::ShapeAwareFunctional< ::serac::H1<(int)1, (int)3> , ::serac::H1<(int)2, (int)3> ( ::serac::H1<(int)2, (int)3> , ::serac::H1<(int)2, (int)3> ), ( ::serac::ExecutionSpace)0> ::trial_spaces" is undefined in device code

livchat recommends making trial_spaces and test_space members of the functor. for now, i removed the SERAC_HOST_DEVICE macros from the operator() functions.

other

removed here but should probably be re-added at some point: 3ba0e94

[chapman39]

need to re-add it to pipeline rule

[chapman39]

functional isn't designed to run on gpu. we need to either compile those files as cxx or not build with functional on cuda builds