Restored recommendations based on acoustic similarities (using musicnn), fixes #301

Dockerfile-build-arch

@@ -12,7 +12,9 @@ ARG BUILD_PACKAGES="\
 	libconfig \
 	libpulse \
 	make \
+	onnxruntime-cpu \
 	pkgconfig \
+	protobuf \
 	pugixml \
 	stb \
 	taglib \
@@ -33,4 +35,4 @@ ARG LMS_UNITY_BUILD=ON
 ARG LMS_IMAGE_BACKEND=stb
 RUN cmake .. -DCMAKE_BUILD_TYPE=${LMS_BUILD_TYPE} -DCMAKE_UNITY_BUILD=${LMS_UNITY_BUILD} -DLMS_IMAGE_BACKEND=${LMS_IMAGE_BACKEND} -DBUILD_BENCHMARKS=ON
 RUN VERBOSE=1 make -j$(nproc)
-RUN make test
+RUN LMS_MUSICNN_MODEL=/tmp/lms/src/libs/audio/models/MSD_musicnn_embedding.onnx make test