Fix add of GPU Fausts.

Fixing the C++ functions behind (TransformHelper<FPP,GPU2>::multiply() and push_back that wasn't forwarding the transpose and conjugate flags to underlying Transform object).

(test_pyfaust-2.9.0) hhadjdji@gtx980:~$ python3 -c "from pyfaust import rand as frand; gpuF = frand(5, [5,10], dev='gpu', fac_type='dense');  cpuF = gpuF.clone(dev='cpu'); print(cpuF+cpuF); print((cpuF+cpuF).norm()); print(gpuF+gpuF);print((gpuF+gpuF).norm())"
loading libgm
Faust size 10x5, density 8.8, nnz_sum 440, 8 factor(s):
- FACTOR 0 (real) SPARSE, size 10x20, density 0.1, nnz 20
- FACTOR 1 (real) SPARSE, size 20x10, density 0.5, nnz 100
- FACTOR 2 (real) SPARSE, size 10x16, density 0.3125, nnz 50
- FACTOR 3 (real) SPARSE, size 16x18, density 0.277778, nnz 80
- FACTOR 4 (real) SPARSE, size 18x16, density 0.3125, nnz 90
- FACTOR 5 (real) SPARSE, size 16x10, density 0.5, nnz 80
- FACTOR 6 (real) SPARSE, size 10x10, density 0.1, nnz 10
- FACTOR 7 (real) SPARSE, size 10x5, density 0.2, nnz 10

338.16702092577
- GPU FACTOR 0 (real) SPARSE size 10 x 20, addr: 0x133a3840, density 0.100000, nnz 20
- GPU FACTOR 1 (real) SPARSE size 20 x 10, addr: 0x133a2950, density 0.500000, nnz 100
- GPU FACTOR 2 (real) SPARSE size 10 x 16, addr: 0x133a1b00, density 0.312500, nnz 50
- GPU FACTOR 3 (real) SPARSE size 16 x 18, addr: 0x133a0cb0, density 0.277778, nnz 80
- GPU FACTOR 4 (real) SPARSE size 18 x 16, addr: 0x133472b0, density 0.312500, nnz 90
- GPU FACTOR 5 (real) SPARSE size 16 x 10, addr: 0xa186360, density 0.500000, nnz 80
- GPU FACTOR 6 (real) SPARSE size 10 x 10, addr: 0xa186ae0, density 0.100000, nnz 10
- GPU FACTOR 7 (real) SPARSE size 10 x 5, addr: 0xa497740, density 0.200000, nnz 10

338.16702092577
k