[PyTorch] Adding support for all possible torch.cuda and c10::cuda APIs & Macros

[TejaX-Alaghari]

This PR adds support for below -

torch.cuda. APIs:

• torch.cuda.init()
• torch.cuda.is_initialized()
• torch.cuda.current_device()
• torch.cuda.set_device(dev)
• torch.cuda.get_device_properties(dev)
• torch.cuda.get_device_name()
• torch.cuda.device_of(obj)
• torch.cuda.synchronize()
• torch.cuda.current_stream()
• torch.cuda.StreamContext(st)
• torch.cuda.stream(st)
• torch.cuda.set_stream(st)

c10::cuda:: APIs:

• c10::cuda::device_count()
• c10::cuda::device_count_ensure_non_zero()
• c10::cuda::current_device()
• c10::cuda::set_device(idx)
• DeviceIndex c10::cuda::ExchangeDevice(idx)
• DeviceIndex c10::cuda::MaybeExchangeDevice(idx)

c10::cuda:: Macros:

• C10_CUDA_BUILD_SHARED_LIBS
• C10_CUDA_EXPORT
• C10_CUDA_IMPORT
• C10_CUDA_API
• C10_CUDA_BUILD_MAIN_LIB