llama.cpp

ggml-base: use MATH_LIBRARY variable instead of hardcoded 'm'

#22239

Merged