llama.cpp

Fix: Handle unaligned writes in ggml_backend_webgpu_buffer_set_tensor

#29471

Merged