Previously, cmake would always build targeting cuda.

We now dispatch on FF_GPU_BACKEND to target either cuda, hip_cuda,
or hip_rocm. The hip backends are expected to use the hip cmake config
file that comes with a hip install. The discrepancies for targeting
hip_cuda vs hip_rocm are documented inline.

Configure legion for hip backend

add legion max return size check

fix std::stringstream var decl parsed as function

explicit template instantiations
include string

We glob for files under src to get the source files for the flexflow
target. Moving tools to the top level directory prevents the tools
sourcefiles from accidentally being added to the flexflow target
source files.

change substitution_to_dot cuda_add_executable to add_executable.
When building with hip_rocm, we don't have cuda available and shouldn't
need to build with it for substitution_to_dot as the target does
not directly build and cuda kernels

do not call sed to manually change config script

clone submodules in docker job

…ead of parsing it adhoc