llama.cpp Release b10842 Adds AMD gfx90c HIP Support

Open-source project llama.cpp released version b10842, introducing HIP support for AMD's gfx90c GPU architecture within the GGML tensor backend. Pull request #26454 ensures that this HIP implementation fully complies with technical specifications. This update expands hardware compatibility for local LLM execution onto integrated or legacy AMD hardware featuring the gfx90c architecture, such as Renoir APUs. It enables users with these graphics chips to utilize GPU acceleration via the GGML ecosystem. The release provides pre-built binaries across multiple operating systems, including Linux and Windows builds compatible with ROCm 10.0, Vulkan, and CUDA backends. The code modifications specifically ensure specification-compliant HIP execution for devices targeted by the `gfx90c` target ISA.