Match Qwen head-chunked attention execution

docs/issue113-head-chunk/deepseek.log

@@ -0,0 +1,10 @@
+
+running 1 test
+test engine::metal::tests::resident_multi_session_switching_preserves_each_kv_frontier ... ds4: Metal device Apple M5 Max, 128.00 GiB RAM
+ds4: Metal 4 tensor API enabled for Tensor kernels
+ds4: drift-patch flags hc_stable=on norm_unify=on kv_raw_f32=off rope_exp2_log2=off math_safe=off tensor_matmul=on
+ds4: Metal model views created in 1.885 ms, residency requested in 9117.538 ms, warmup 3.450 ms (mapped 82697.67 MiB from offset 5.09 MiB)
+ds4: Metal mapped mmaped model as 2 overlapping shared buffers
+ok
+
+test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 548 filtered out; finished in 9.69s