Wait for an NvSciSyncFence.

Use cudaWaitExternalSemaphoresAsync to make all operations enqueued on the CUDA stream (passed as a parameter to this API) wait until the NvSciSyncFence (sent as a parameter to this API via paramsArray) is signaled by the relevant signaler. Such a wait happens asynchronously on the GPU (i.e., the calling thread returns immediately). Applications can also optionally set flag CUDA_EXTERNAL_SEMAPHORE_WAIT_SKIP_NVSCIBUF_MEMSYNC to indicate that memory synchronization operations are disabled over all CUDA-NvSciBufs imported into CUDA (in that process), which are normally performed by default to ensure data coherency with other importers of the same NvSciBuf memory objects. Use this flag when CUDA-NvSciSync is used to build only control-dependencies (i.e., no data sharing between the signaler and waiter).