Start the inference engine from both apps and show its choice in Settings

The desktop names where a package may have put libonnxruntime — an
override variable, beside the executable, the package's own library
directory, the Flatpak prefix, the system library directory — and
Android points at the APK's native library directory, which is also
what Qualcomm's DSP loader must be told for the Hexagon skel. Android
starts the engine at the end of the model unpack rather than at launch,
because the probe fingerprints the model files and a first launch has
none until then.

The About panel gains an Inference row beside Graphics, re-read every
two seconds while the probe runs and engines land, and faces.model_id
carries the detector's form: an int8 detector finds a different set of
faces and is a different population (docs/inference.md §7). A
low-memory signal drops every idle session with the GPU caches.

The APK assembly bundles ONNX Runtime and the Qualcomm HTP libraries
from Maven, fetched by tools/fetch-android-runtime.sh with their
published checksums; RUNTIME_DIR=none builds the tract-only APK, which
is a slower app and not a broken one. The desktop packages carry no
runtime yet.

Two probe fixes from the first desktop run: the floor must not be
built with CPU fallback disabled, and a versioned libonnxruntime.so is
a runtime too. On the reference desktop the probe now loads ONNX
Runtime 1.30, measures 30 ms on the CPU provider, and selects TensorRT
at 1.5 ms.

docker/android/assemble-apk.sh

@@ -17,6 +17,8 @@
 #   JNILIBS      where cargo-ndk wrote the .so (default: $TARGET_DIR/jniLibs)
 #   OUT          output directory             (default: $TARGET_DIR/apk)
 #   KEYSTORE     signing keystore             (default: $TARGET_DIR/debug.keystore)
+#   RUNTIME_DIR  the inference runtime         (default: $TARGET_DIR/runtime,
+#                                               fetched by tools/fetch-android-runtime.sh)
 #   ABI          Android ABI                  (default: arm64-v8a)
 #   RUST_TARGET  Rust target triple           (default: aarch64-linux-android)
 #
@@ -43,6 +45,7 @@ TARGET_DIR="$(cd "${TARGET_DIR}" && pwd)"
 JNILIBS="${JNILIBS:-${TARGET_DIR}/jniLibs}"
 OUT="${OUT:-${TARGET_DIR}/apk}"
 KEYSTORE="${KEYSTORE:-${TARGET_DIR}/debug.keystore}"
+RUNTIME_DIR="${RUNTIME_DIR:-${TARGET_DIR}/runtime}"
 
 # Release signing is selected by supplying a password, not by a flag, so there
 # is no way to ask for a release build and silently get a debug one.
@@ -255,6 +258,25 @@ fi
 cp "${SO}" "${OUT}/staging/lib/${ABI}/libdarkroom.so"
 cp "${DEX}" "${OUT}/staging/classes.dex"
 
+# The inference runtime (docs/inference.md §3): ONNX Runtime and Qualcomm's
+# Hexagon backend, beside libdarkroom.so so the app finds them in its own
+# native library directory. The build links none of it — the app dlopens
+# `libonnxruntime.so` at launch and runs on tract if it is not there — so an
+# APK without these is a slower app, not a broken one, and `RUNTIME_DIR=none`
+# builds exactly that. 174 MB for the default set; the script says which
+# Hexagon generations that buys.
+if [[ "${RUNTIME_DIR}" != "none" ]]; then
+    if [[ ! -f "${RUNTIME_DIR}/lib/libonnxruntime.so" ]]; then
+        "${REPO}/tools/fetch-android-runtime.sh" "${RUNTIME_DIR}"
+    fi
+    cp "${RUNTIME_DIR}"/lib/*.so "${OUT}/staging/lib/${ABI}/"
+    mkdir -p "${OUT}/staging/assets/licences"
+    cp "${RUNTIME_DIR}"/QNN-*.* "${OUT}/staging/assets/licences/" 2>/dev/null || true
+    echo "    runtime: $(ls "${RUNTIME_DIR}/lib" | wc -l) libraries from ${RUNTIME_DIR}/lib"
+else
+    echo "    runtime: none (tract only)"
+fi
+
 # The models. Android has no other route to one — app-private storage is not
 # user-reachable and the in-app fetch is unbuilt (docs/faces.md §2.2a) — so
 # they go in the APK and `android_main` unpacks them on first launch. The
@@ -274,7 +296,7 @@ cp "${DEX}" "${OUT}/staging/classes.dex"
 #
 # Cleared first: a previous run that died between staging and cleanup would
 # otherwise leave models in the APK that are no longer in the tree.
-rm -rf "${OUT}/staging/assets"
+rm -rf "${OUT}/staging/assets/models"
 mkdir -p "${OUT}/staging/assets/models"
 _bundled=""
 for _dir in face scene; do
@@ -315,7 +337,7 @@ fi
 # install times sane.
 cd "${OUT}/staging"
 cp "${OUT}/base.apk" "${OUT}/unaligned.apk"
-zip -q -0 -X "${OUT}/unaligned.apk" "lib/${ABI}/libdarkroom.so"
+zip -q -0 -X "${OUT}/unaligned.apk" lib/"${ABI}"/*.so
 zip -q -X "${OUT}/unaligned.apk" classes.dex
 # Stored, not deflated: an ONNX graph is mostly incompressible float data, so
 # deflating it buys a few percent and costs the whole file being inflated into