Allow HIP to be used with NVIDIA GPUs

[davebayer]

Currently when HIP as the library backend, vendor and other parameters are set to AMD. We can use __HIP_PLATFORM_AMD__ and __HIP_PLATFORM_NVIDIA__ macros to detect what is the target device to set up everything properly.

[DTolm]

Hello,

Sorry for the long reply, I am busy with another project at this moment. I can add subflags for HIP version of Nvidia (if anyone even uses it), however, the solution you provided in the pull request is not correct. All the parameters in the backend-specific parts were specifically tuned for the particular backend, which is why they are not unified.

Best regards, Dmitrii