Support using onnxruntime 1.16.0 with CUDA 11.4 on Jetson Orin NX.

The pre-built onnxruntime libs is provided by the community
using the following command:

```
./build.sh --build_shared_lib --config Release --update \
  --build --parallel --use_cuda \
  --cuda_home /usr/local/cuda \
  --cudnn_home /usr/lib/aarch64-linux-gnu 2>&1 | tee my-log.txt
```

See also microsoft/onnxruntime#11226

CMakeLists.txt

@@ -46,6 +46,9 @@ option(SHERPA_ONNX_USE_PRE_INSTALLED_ONNXRUNTIME_IF_AVAILABLE "True to use pre-i
 option(SHERPA_ONNX_ENABLE_SANITIZER "Whether to enable ubsan and asan" OFF)
 option(SHERPA_ONNX_BUILD_C_API_EXAMPLES "Whether to enable C API examples" ON)
 
+set(SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION "1.11.0" CACHE STRING "Used only for Linux ARM64 GPU. If you use Jetson nano b01, then please set it to 1.11.0. If you use Jetson Orin NX, then set it to 1.16.0")
+
+
 set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/lib")
 set(CMAKE_LIBRARY_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/lib")
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/bin")

cmake/onnxruntime-linux-aarch64-gpu.cmake

@@ -18,19 +18,37 @@ if(NOT SHERPA_ONNX_ENABLE_GPU)
   message(FATAL_ERROR "This file is for NVIDIA GPU only. Given SHERPA_ONNX_ENABLE_GPU: ${SHERPA_ONNX_ENABLE_GPU}")
 endif()
 
-set(onnxruntime_URL  "https://github.com/csukuangfj/onnxruntime-libs/releases/download/v1.11.0/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2")
-set(onnxruntime_URL2 "https://hf-mirror.com/csukuangfj/onnxruntime-libs/resolve/main/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2")
-set(onnxruntime_HASH "SHA256=36eded935551e23aead09d4173bdf0bd1e7b01fdec15d77f97d6e34029aa60d7")
+message(WARNING "\
+SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION: ${SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION}
+If you use Jetson nano b01, then please pass
+   -DSHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION=1.11.0
+to cmake (You need to make sure CUDA 10.2 is available on your board).
+
+If you use Jetson Orin NX, then please pass
+   -DSHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION=1.16.0
+to cmake (You need to make sure CUDA 11.4 is available on your board).
+")
+
+set(_v ${SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION})
+
+set(onnxruntime_URL  "https://github.com/csukuangfj/onnxruntime-libs/releases/download/v$v/onnxruntime-linux-aarch64-gpu-$v.tar.bz2")
+set(onnxruntime_URL2 "https://hf-mirror.com/csukuangfj/onnxruntime-libs/resolve/main/onnxruntime-linux-aarch64-gpu-$v.tar.bz2")
+
+if(_v STREQUAL "1.11.0")
+  set(onnxruntime_HASH "SHA256=36eded935551e23aead09d4173bdf0bd1e7b01fdec15d77f97d6e34029aa60d7")
+else()
+  set(onnxruntime_HASH "SHA256=6e671150318fb3634f09be8999236b3de6178e949cd91dc6c2c8a056c081cfdb")
+endif()
 
 # If you don't have access to the Internet,
 # please download onnxruntime to one of the following locations.
 # You can add more if you want.
 set(possible_file_locations
-  $ENV{HOME}/Downloads/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2
-  ${CMAKE_SOURCE_DIR}/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2
-  ${CMAKE_BINARY_DIR}/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2
-  /tmp/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2
-  /star-fj/fangjun/download/github/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2
+  $ENV{HOME}/Downloads/onnxruntime-linux-aarch64-gpu-$v.tar.bz2
+  ${CMAKE_SOURCE_DIR}/onnxruntime-linux-aarch64-gpu-$v.tar.bz2
+  ${CMAKE_BINARY_DIR}/onnxruntime-linux-aarch64-gpu-$v.tar.bz2
+  /tmp/onnxruntime-linux-aarch64-gpu-$v.tar.bz2
+  /star-fj/fangjun/download/github/onnxruntime-linux-aarch64-gpu-$v.tar.bz2
 )
 
 foreach(f IN LISTS possible_file_locations)