Need CUDA infastructure

[ye-luo]

Needed by #239.
The CUDA infrastructure will be need.
One way of doing this is copying everything from QMCPACK.
However, I noticed one_code branch has some CUDA infrastructure as well. @PDoakORNL could you comment what we can learn from one_code? Do you want to bring your CUDA setting over to miniQMC develop?

I tried one_code branch and noticed that it seems support Clang CUDA. I think it is a useful feature for us not just depending on nvcc. I also encounter an issue that LDFLAGS is required to specify paths to cublas and cusolve. This is not convenient for users. Maybe this is just an minor issue.