llama.cpp b10037 allows manual tensor types with --pure

The llama.cpp project has released version b10037, which includes a change to allow the use of manual tensor types when using the --pure flag.

• macOS Apple Silicon (arm64) binaries are available.
• macOS Intel (x64) binaries are provided.
• iOS XCFramework is included in the release.
• Ubuntu builds cover CPU, Vulkan, ROCm 7.2, OpenVINO, and SYCL targets for x64, arm64, and s390x architectures.
• Android arm64 (CPU) binaries are released.
• Windows builds support CPU, OpenCL Adreno, CUDA 12/13, Vulkan, OpenVINO, SYCL, and HIP for x64 and arm64.
• openEuler builds for x86 and aarch64 architectures are available.
• A standalone UI package is also distributed.