llama.cpp b9987 adds gguf tensor shape accessors

The llama.cpp project released version b9987, which introduces new tensor shape accessors to the GGUF format. This update modifies the API by returning tensor shapes as const int64_t pointers and removes the n_dims accessor in favor of gguf_get_tensor_ne.

• macOS/iOS binaries for Apple Silicon (arm64) and Intel (x64), plus an iOS XCFramework.
• Linux builds for Ubuntu x64, arm64, and s390x CPUs, along with Vulkan, ROCm 7.2, OpenVINO, and SYCL variants.
• Android arm64 CPU binaries.
• Windows binaries for CPU, OpenCL Adreno, CUDA 12/13, Vulkan, OpenVINO, SYCL, and HIP.
• openEuler support for x86 and aarch64 architectures with ACL Graph.

This release provides updated binaries across major platforms and hardware accelerators while refining the internal GGUF tensor access interface.