llama.cpp b10189 Removes Custom CPU Op for M3 Graph

The latest release of llama.cpp (b10189) removes a custom CPU operation from the M3 graph, replacing it with standard operations. This change ensures better compatibility and performance across different hardware configurations without relying on proprietary optimizations. Operators running local AI workloads should be aware that this update may affect models using the M3 architecture, particularly those optimized for specific CPUs. The removal of custom ops also simplifies maintenance and future updates by aligning more closely with industry-standard practices.