Android use new llm runner deps (pytorch#14361)

Use extension_llm_runner instead of old llama_runner llava_runner

Author: kirklandsign

examples/qualcomm/oss_scripts/llama/runner/runner.cpp

@@ -182,8 +182,7 @@ Error Runner<T>::load() {
     eos_ids->insert(tokenizer_->encode("<|eot|>", 0, 0).get()[0]);
     eos_ids->insert(tokenizer_->encode("<|end_of_text|>", 0, 0).get()[0]);
   } else {
-    tokenizer_ =
-        example::load_llama_tokenizer(tokenizer_path_, Version::Default);
+    tokenizer_ = llm::load_tokenizer(tokenizer_path_);
     if (tokenizer_ == nullptr) {
       ET_LOG(
           Error, "Failed to load tokenizer with %s", tokenizer_path_.c_str());

extension/android/CMakeLists.txt

@@ -168,21 +168,8 @@ endif()
 
 if(EXECUTORCH_BUILD_LLAMA_JNI)
   target_sources(executorch_jni PRIVATE jni/jni_layer_llama.cpp jni/log.cpp)
-  list(APPEND link_libraries llama_runner)
+  list(APPEND link_libraries extension_llm_runner)
   target_compile_definitions(executorch_jni PUBLIC EXECUTORCH_BUILD_LLAMA_JNI=1)
-  add_subdirectory(
-    ${EXECUTORCH_ROOT}/examples/models/llama/runner
-    ${CMAKE_CURRENT_BINARY_DIR}/../../examples/models/llama/runner
-  )
-
-  target_sources(
-    executorch_jni
-    PRIVATE ${EXECUTORCH_ROOT}/extension/llm/runner/llm_runner_helper.cpp
-  )
-
-  target_include_directories(
-    executorch_jni PRIVATE ${EXECUTORCH_ROOT}/extension/llm/runner
-  )
 
   if(QNN_SDK_ROOT)
     target_sources(