Function quantize_activations_q8_k 

pub fn quantize_activations_q8_k(x: &[f32]) -> Q8KActivations

Quantize activations to ggml Q8_K (256-elem super-blocks). Positive scale convention (d = amax/127) matching our Q8_0 path; bsums enable the Q4_K min correction without re-scanning q.