Streamlines don't compile with cuda unless special flags are added
These flags have been added in the testing and an example, but there is no way for downstream libs to know anything about this flags since streamlines don't seem to be compiled into the filter lib.
nvlink error : Entry function '_ZN4vtkm4cont4cuda8internal19TaskStrided1DLaunchINS_4exec4cuda8internal13TaskStrided1DIKNS_7worklet17particleadvection21ParticleAdvectWorkletEKNS_8internal10InvocationINSC_17FunctionInterfaceIFvNSC_19ArrayPortalImplicitINSC_12IndexFunctorEEEPKNS9_14IntegratorBase10ExecObjectENS9_37StateRecordingParticleExecutionObjectINS_8ParticleEEENSF_INS0_8internal15ConstantFunctorIiEEEEEEENSE_IFvNS8_15WorkletMapField7FieldInENS8_8internal11WorkletBase10ExecObjectESZ_SW_EEENSE_IFvNS_12placeholders3ArgILi1EEENS13_ILi2EEENS13_ILi3EEENS13_ILi4EEEEEELi1ESH_SS_SH_NS0_20DeviceAdapterTagCudaEEEEEEEvT_i' uses too much data for compiler-generated constants; please recompile with -Xptxas --disable-optimizer-constants