[Feature] DirectML execution provider for Machine Learning (AMD GPU Support on Windows/WSL2)

I have searched the existing feature requests, both open and closed, to make sure this is not a duplicate request. • Yes The feature Hi Immich team! First of all, thank you for the amazing work on this project. I would like to request an official Machine Learning container image that includes the DirectML execution provider for ONNX Runtime. Current Limitation: Currently, users running Immich on Windows via Docker Desktop (WSL2) with AMD GPUs cannot utilize hardware acceleration for the ML container. NVIDIA users can use the -cuda image (via WSL2 GPU paravirtualization). Intel users can use the -openvino image (via /dev/dxg). AMD users are instructed to use -rocm, which requires native Linux kernel paths (/dev/kfd and /dev/dri) that are not passed through or supported in WSL2. Because of this, AMD GPU users are currently forced to use CPU-only processing on Windows. Proposed Solution: Since Microsoft's DirectML acts as a hardware-agnostic layer over D…