add FastCaloSim-GPU

Use GPU to accelerate the Fast Calorimeter simulation. To use it, first enable the GPU in compiling (cmake -DUSE_GPU=ON ...), then change the flag (RunOnGPU) in FastCaloSimV2ParamSvc to True. Please be noted the cuda (>=11.3) is needed in compiling and if no GPU device is found in the running it will use CPU. A series of tests have been conducted for single particle and ttbar events and the results looks good.

Ping @jchapman , @leggett , @tlari , @mfauccig and @cyoung .