Support building GPU-capable sherpa-onnx on Linux aarch64. (#1500)

Thanks to @Peakyxh for providing pre-built onnxruntime libraries 
with CUDA support for Linux aarch64.

Tested on Jetson nano b01

sherpa-onnx/csrc/speaker-embedding-extractor-nemo-model.cc

@@ -53,7 +53,7 @@ class SpeakerEmbeddingExtractorNeMoModel::Impl {
     return std::move(outputs[0]);
   }
 
-  OrtAllocator *Allocator() const { return allocator_; }
+  OrtAllocator *Allocator() { return allocator_; }
 
   const SpeakerEmbeddingExtractorNeMoModelMetaData &GetMetaData() const {
     return meta_data_;