feat: GPU acceleration support for ONNX Runtime inference

Centralize execution provider selection in providers.py with
auto-detection and graceful fallback. All ONNX sessions (embedder,
detector, face processor, recognizer) now use the configured providers.

- New VISION_EXECUTION_PROVIDERS env var: "auto" for GPU auto-detect,
  or explicit "CUDAExecutionProvider,CPUExecutionProvider"
- Provider priority: CUDA > ROCm > OpenVINO > CPU (when set to "auto")
- docker-compose.yml includes commented-out NVIDIA GPU deploy section
- Supports onnxruntime-gpu as a drop-in replacement for onnxruntime

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

docker-compose.yml

@@ -143,6 +143,13 @@ services:
       - LOG_LEVEL=${LOG_LEVEL:-INFO}
       - TZ=${TZ:-UTC}
       - MULITA_CELERY_WORKER=1
+      # ONNX Runtime execution providers. Set to "auto" to auto-detect
+      # GPU (CUDA > ROCm > OpenVINO > CPU), or explicitly:
+      #   "CUDAExecutionProvider,CPUExecutionProvider"
+      #   "ROCMExecutionProvider,CPUExecutionProvider"
+      # Default: CPU only. To enable GPU, also uncomment the deploy
+      # section below and install nvidia-container-toolkit on the host.
+      - VISION_EXECUTION_PROVIDERS=${VISION_EXECUTION_PROVIDERS:-CPUExecutionProvider}
       # Pin each ONNX session to one intra-op thread so N prefork children
       # × default-all-cores doesn't oversubscribe the box. With
       # concurrency=5 and OMP=1, vision peaks at 5 busy cores, leaving
@@ -151,6 +158,14 @@ services:
       - OMP_NUM_THREADS=1
       - OPENBLAS_NUM_THREADS=1
       - MKL_NUM_THREADS=1
+    # Uncomment for NVIDIA GPU passthrough:
+    # deploy:
+    #   resources:
+    #     reservations:
+    #       devices:
+    #         - driver: nvidia
+    #           count: all
+    #           capabilities: [gpu]
     depends_on:
       redis:
         condition: service_started