[Feat] Unquantized Linear to nz and control all nz-cast (#3356)

### What this PR does / why we need it?
Currently, when executing to the Linear layer of models in vLLM-Ascend,
the weights format is ND in unquantized case and skipped ascend case.
This PR supplements the execution logic for Linear layer. We use a new
global variable: VLLM_ASCEND_ENABLE_NZ. When VLLM_ASCEND_ENABLE_NZ=1 and
CANN version is 8.3, the weights of the Linear layer will be converted
to FRACTAL_NZ, in both unquantized case and skipped ascend case. We also
use VLLM_ASCEND_ENABLE_NZ to control the existing NZ conversion, such as
w8a8-quantized case.

### Does this PR introduce _any_ user-facing change?
Add a new global variable VLLM_ASCEND_ENABLE_NZ. If you want to use NZ
format, you should set VLLM_ASCEND_ENABLE_NZ=1.

### 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: anon189Ty <Stari_Falcon@outlook.com>

vllm_ascend/worker/model_runner_v1.py

@@ -97,6 +97,7 @@ from vllm.v1.worker.utils import (AttentionGroup, bind_kv_cache,
                                   sanity_check_mm_encoder_outputs,
                                   scatter_mm_placeholders)
 
+import vllm_ascend.envs as envs_ascend
 from vllm_ascend.ascend_config import get_ascend_config
 from vllm_ascend.ascend_forward_context import (MoECommType,
                                                 set_ascend_forward_context)
@@ -125,7 +126,7 @@ from vllm_ascend.spec_decode.interface import SpecDcodeType
 from vllm_ascend.spec_decode.mtp_proposer import MtpProposer
 from vllm_ascend.utils import (ACL_FORMAT_FRACTAL_ND, ACL_FORMAT_FRACTAL_NZ,
                                AscendSocVersion, ProfileExecuteDuration,
-                               get_ascend_soc_version, is_310p,
+                               get_ascend_soc_version, is_310p, is_enable_nz,
                                lmhead_tp_enable)
 from vllm_ascend.worker.npu_input_batch import CachedRequestState, InputBatch
 
@@ -137,8 +138,6 @@ else:
 
 import torch_npu
 
-import vllm_ascend.envs as envs_ascend
-
 # if true, allow tensor initialization and casting with internal format (e.g., NZ)
 torch.npu.config.allow_internal_format = True
 
@@ -2609,6 +2608,9 @@ class NPUModelRunner(LoRAModelRunnerMixin):
                                          runtime_mode=CUDAGraphMode.FULL)
 
     def _convert_torch_format(self, tensor):
+        if ACL_FORMAT == ACL_FORMAT_FRACTAL_NZ \
+                and not is_enable_nz():
+            return tensor
         tensor = torch_npu.npu_format_cast(tensor, ACL_FORMAT)
         return tensor