feat: add GPU/CUDA acceleration support (:cuda Docker tag)

Add a :cuda Docker image tag that auto-detects NVIDIA GPU at runtime
and falls back gracefully to CPU. Same pattern as Immich.

- New gpu.py shared utility for cached CUDA detection
- Background removal (rembg): pass CUDAExecutionProvider to ONNX Runtime
- Upscaling (Real-ESRGAN): use CUDA device + FP16 when GPU available
- OCR (PaddleOCR): enable use_gpu when CUDA detected
- Dispatcher reports GPU status at startup via readiness signal
- Admin health endpoint exposes GPU availability
- Dockerfile uses ARG GPU=false with conditional NVIDIA CUDA base image
- docker-compose.gpu.yml override for GPU users
- CI/CD workflows build and publish :cuda tag (amd64 only)

Three tags: :latest (CPU), :lite (no AI), :cuda (GPU with CPU fallback)

packages/ai/src/bridge.ts

@@ -54,6 +54,8 @@ interface PendingRequest {
 let dispatcher: ChildProcess | null = null;
 let dispatcherReady = false;
 let dispatcherFailed = false;
+// biome-ignore lint/style/useConst: reassigned on dispatcher readiness signal
+let dispatcherGpuAvailable = false;
 const pendingRequests = new Map<string, PendingRequest>();
 let stdoutBuffer = "";
 
@@ -82,6 +84,7 @@ function startDispatcher(): ChildProcess | null {
           // Readiness signal
           if (parsed.ready === true) {
             dispatcherReady = true;
+            dispatcherGpuAvailable = parsed.gpu === true;
             continue;
           }
 
@@ -221,6 +224,13 @@ function dispatcherRun(
   });
 }
 
+/**
+ * Whether the Python dispatcher detected a CUDA GPU at startup.
+ */
+export function isGpuAvailable(): boolean {
+  return dispatcherGpuAvailable;
+}
+
 /**
  * Shut down the persistent dispatcher process.
  */

packages/ai/src/index.ts

@@ -1,5 +1,5 @@
 export { removeBackground } from "./background-removal.js";
-export { shutdownDispatcher } from "./bridge.js";
+export { isGpuAvailable, shutdownDispatcher } from "./bridge.js";
 export { blurFaces } from "./face-detection.js";
 export { inpaint } from "./inpainting.js";
 export { extractText } from "./ocr.js";