[BugFix] Fix the bug when enabling both DBO and MoE multistream of deepseek. (#1759)

When enabling both DBO and MoE multistream of deepseek model, there
exists problem. This PR fixes this situation and add e2e ut test for
this scenario.

Signed-off-by: whx-sjtu <2952154980@qq.com>

Author: whx-sjtu

vllm_ascend/models/deepseek_dbo.py

@@ -149,41 +149,6 @@ def __init__(
             )
         CustomDeepseekDBOMoE.top_k = config.num_experts_per_tok
 
-    def forward(
-            self,
-            hidden_states: torch.Tensor,
-            attn_metadata: Optional[AttentionMetadata] = None) -> torch.Tensor:
-        forward_context = get_forward_context()
-        if attn_metadata is None:
-            attn_metadata = forward_context.attn_metadata
-
-        # when profile runs, force experts to load balanced tokens
-        # to avoid high memory consumption on a single rank.
-        enable_force_load_balance = forward_context.in_profile_run
-
-        is_prefill = forward_context.with_prefill
-        # If this node is kv_consumer, we force the moe always runs in decode path to make sure
-        # the behaviour aligned between dummy_run and normal model_execute.
-        if self.kv_consumer:
-            is_prefill = False
-
-        # router_logits: (num_tokens, n_experts)
-        router_logits, _ = self.gate(hidden_states)
-
-        experts_hidden_states = self.experts(
-            hidden_states=hidden_states,
-            router_logits=router_logits,
-            is_prefill=is_prefill,
-            top_k=CustomDeepseekDBOMoE.top_k,
-            enable_force_load_balance=enable_force_load_balance,
-            shared_experts=self.shared_experts)
-
-        hidden_states = (
-            experts_hidden_states[0] * self.routed_scaling_factor +
-            experts_hidden_states[1])
-
-        return hidden_states
-
     # ----------------------------------------- TBO-related --------------------------------------------
     def _forward_ms_op_shared_expert(
         self,