gen - add 3D mixed support

cuda - fix 2D flattening

gpu - use 2d Flat variants in gen

gpu - isolate core 2D tensor logic to allow flat version

gpu - add P_1D to template args for AtPoints

gen - use blocksize of 1 elem AtPoints

cuda - use BASIS_T_1D in codegen

gen - restrict input/output array pointers

gen - use field names for clarity

gpu - gen should use GetArray over GetArrayWrite

gpu - gen put suboperators on separate streams

gpu - gen ApplyAdd functions

gpu - allow running shared kernels on stream

gpu - isolate gen ApplyAdd inner logic

gpu - use struct over array for clarity

cuda - remove duplicate mats in gen

gpu - swap out bitwise assignment operators for bools

Co-authored-by: Zach Atkins <zach.atkins@colorado.edu>

hip - gen fallback to shared if error

cuda - gen fallback to shared if error

gpu - update gen non-tensor block strategy

hip - nontensor gen operators

cuda - nontensor gen operators

hip - AtPoints for hip/gen

cuda - AtPoints for cuda/gen

gen - fun name standardization