[CUDA] Initial implementation of Convolution with cuDNN (#2385)

* Link with cuDNN

* Initial implementation

* Remove backend apis

* Fix recording cudnn conv

* More unused backend apis

* Fix C++ conv tests

* include cudnn as python dep

* Install libcudnn9-dev-cuda-12 in CI

* cudnn only accepts contiguous inputs

* Switch to backend apis

* Plan needs to be kept alive

* Turn off tf32

* Add cache

* Test the native cuda graph api

* Set cudnn stream before execution

* Make LRUCache more like a normal container

* Do error check for cublas handle

* Zero-initilizing array

* Use tf32 for conv

* Skip TestConv.test_torch_conv_2D test

---------

Co-authored-by: Awni Hannun <awni@apple.com>

mlx/backend/cuda/utils.cpp

@@ -17,6 +17,14 @@ CudaStream::~CudaStream() {
   CHECK_CUDA_ERROR(cudaStreamDestroy(stream_));
 }
 
+void check_cublas_error(const char* name, cublasStatus_t err) {
+  if (err != CUBLAS_STATUS_SUCCESS) {
+    // TODO: Use cublasGetStatusString when it is widely available.
+    throw std::runtime_error(
+        fmt::format("{} failed with code: {}.", name, static_cast<int>(err)));
+  }
+}
+
 void check_cuda_error(const char* name, cudaError_t err) {
   if (err != cudaSuccess) {
     throw std::runtime_error(