Add automatic warp count control for low-latency kernels (#213)

* Add automatic warp count control for low-latency dispatch

* Add automatic warp count control for low-latency combine

* More assertions

csrc/deep_ep.hpp

@@ -41,6 +41,7 @@ private:
 
     // Device info and communication
     int device_id;
+    int num_device_sms;
     int rank, rdma_rank, nvl_rank;
     int num_ranks, num_rdma_ranks, num_nvl_ranks;
     cudaIpcMemHandle_t ipc_handles[NUM_MAX_NVL_PEERS];