Merge Q, K, V (#878)

* POC: merge Q, K, V into a single, contiguous tensor

Done just for Qwen3-MoE, where I see a 4% uplift in TG.
PP performance gain is sub-percent, if any.
Still, it seems it makes sense to do it in general given
the TG performance gain.

* WIP

* merge_qkv: it works for gpt-oss

...but we see a smaller TG gain (~1.5%)

* WIP

* Don't ignore the return value of create_tensors()

else, when q, k, v get merged and we are running on the CPU,
we get a crash because the backend is trying to use mmap,
but that no longer works.

* merge_qkv: bias can be required, optional, or mandatory

* merge_qkv: glm4.5moe

* merge_qkv: add command loine argument to enable

* merge_qkv: fix tensor dimensions

* merge_qkv: llama-4

* merge_qkv: qwen3 (dense)

* merge_qkv: simplify build_qwen3moe

* cohere2 - simplify graph building

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>

src/llama-quantize.cpp

@@ -1007,7 +1007,8 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
         auto v = (std::vector<llama_model_kv_override>*)params->kv_overrides;
         kv_overrides = v->data();
     }
-    llama_model_loader ml(fname_inp, use_mmap, /*check_tensors*/ true, /* repack_tensors */ false, /* use_thp */ false, kv_overrides, nullptr);
+    llama_model_loader ml(fname_inp, use_mmap, /*check_tensors*/ true, /* repack_tensors */ false,
+            /* use_thp */ false, /* merge_qkv */ false, kv_overrides, nullptr);
     ml.init_mappings(false); // no prefetching
 
     llama_model model;