llama-server followups (#16353)

* llama-server followups

Misc fixes for #16031
- Add back dropped ROCm build flag for multi-GPU support on windows
- Fix amdhip64_*.dll version detection for "latest" selection
- Fix embeddings API for consistent normalize behavior with prior versions

* ci: set up for automated llama.cpp update testing

* reduce batch for fa-disabled, and constrained vram

* mlx: fix v3 load bug on m5

Imagegen was incorrectly loading v3 first.  This DRYs out the loading code so imagegen gets the same new v4/v3 selection logic.

* fix reload bug on embedding models

* bump version

* steer user how to enable iGPU when disabled

llama/server/CMakeLists.txt

@@ -96,6 +96,9 @@ if(GGML_HIP AND OLLAMA_RUNNER_DIR MATCHES "^rocm_v")
     ollama_set_cache_default(CMAKE_HIP_FLAGS STRING
         "-parallel-jobs=4" "HIP compiler flags")
     if(WIN32)
+        # Windows ROCm split-load needs peer copies disabled for correctness.
+        ollama_set_cache_default(GGML_CUDA_NO_PEER_COPY BOOL ON
+            "Disable direct peer device copies")
         # HIP on Windows currently emits attributes and deprecated pragma
         # warnings from ROCm headers. Keep the workaround local to Windows ROCm
         # so it can be removed when the ROCm toolchain no longer needs it.