[2/N][Feat] Attention and MoE weight prefetch in Qwen3MoE models (#3203)

### What this PR does / why we need it?

- Refacotr and integrate a unified `WeightPrefetchMethod`
- Integrate `gate_up_proj.weight` in quantized Attention modules
- Prefetching these weights ahead of matmul-like operators imporves
performance by reducing L2 cache transfer latency

### Does this PR introduce _any_ user-facing change?

Add a new config in `--additional-config` for configuration:
```json
{
    "weight_prefetch_config": {
        "enabled": True,
        "prefetch_ratio": {
            "moe": {
                "gate_up": 0.8
            },
        },
    },
}
```
This feature is enabled by default, and can be disabled through this
configuration

### How was this patch tested?


- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: yuzhup <15705211260@163.com>

vllm_ascend/ascend_forward_context.py

@@ -145,7 +145,7 @@ def set_ascend_forward_context(
             forward_context.prefetch_mlp_gate_up_proj = False
             forward_context.prefetch_mlp_down_proj = False
         forward_context.prefetch_mlp_enabled = prefetch_mlp_enabled
-        # TODO(yuzhup): integrate moe weight prefetch method
+        forward_context.model_instance = model_instance
         forward_context.weight_prefetch_method = weight_prefetch_method
 
         # TODO(rjg-lyh): The current implementation is somewhat brute force and not elegant.