Add tiktoken support in setup-with-qnn.sh (#4991)

Adding flags in the build process for QNN backend to enable tiktoken for llama3.

Author: WuhanMonkey

examples/demo-apps/android/LlamaDemo/setup-with-qnn.sh

@@ -34,6 +34,7 @@ cmake examples/models/llama2 \
          -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK/build/cmake/android.toolchain.cmake \
          -DANDROID_ABI="$ANDROID_ABI" \
          -DCMAKE_INSTALL_PREFIX="${CMAKE_OUT}" \
+         -DEXECUTORCH_USE_TIKTOKEN="${EXECUTORCH_USE_TIKTOKEN}" \
          -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \
          -DCMAKE_BUILD_TYPE=Release \
          -B"${CMAKE_OUT}"/examples/models/llama2
@@ -45,6 +46,7 @@ cmake extension/android \
   -DANDROID_ABI="${ANDROID_ABI}" \
   -DCMAKE_INSTALL_PREFIX="${CMAKE_OUT}" \
   -DEXECUTORCH_BUILD_LLAMA_JNI=ON \
+  -DEXECUTORCH_USE_TIKTOKEN="${EXECUTORCH_USE_TIKTOKEN}" \
   -DCMAKE_BUILD_TYPE=Release \
   -B"${CMAKE_OUT}"/extension/android