[bugfix](CP) Fix and unify the PD request discrimination logic. (#5939)

### What this PR does / why we need it?
Since the PR (https://github.com/vllm-project/vllm/pull/32118) has
modified the criteria for judging Prefill and Decode requests in vLLM,
PCPManager needs to synchronize with this standard. As PCPManager
involves multiple calculations of PD request counts, this PR attempts to
consolidate the related logic and update the PD request count once per
batch.

### How was this patch tested?
```bash
pytest tests/e2e/multicard/4-cards/long_sequence/test_mtp.py
```

- vLLM version: v0.13.0
- vLLM main:
11b6af5280

Signed-off-by: QiuChunshuo <qiuchunshuo@huawei.com>

tests/e2e/multicard/4-cards/long_sequence/test_mtp.py

@@ -20,17 +20,18 @@
 import os
 import pytest
 
-from tests.e2e.conftest import VllmRunner
+from tests.e2e.conftest import VllmRunner, wait_until_npu_memory_free
 
 os.environ["HCCL_BUFFSIZE"] = "512"
 
+prompts = [
+    "The capital of France is", "Hello, my name is Tom, I am",
+    "The president of United States is", "AI future is"
+]
+model = "wemaster/deepseek_mtp_main_random_bf16"
 
+@wait_until_npu_memory_free()
 def test_pcp_dcp_mtp1_eager():
-    prompts = [
-        "The capital of France is", "Hello, my name is Tom, I am",
-        "The president of United States is", "AI future is"
-    ]
-    model = "wemaster/deepseek_mtp_main_random_bf16"
     with VllmRunner(
             model,
             max_model_len=1024,
@@ -50,15 +51,8 @@ def test_pcp_dcp_mtp1_eager():
         runner.generate_greedy(prompts, 32)
 
 
-@pytest.mark.skip(
-    reason="vLLM PR-32118 break this",
-)
+@wait_until_npu_memory_free()
 def test_pcp_dcp_mtp3_eager():
-    prompts = [
-        "The capital of France is", "Hello, my name is Tom, I am",
-        "The president of United States is", "AI future is"
-    ]
-    model = "wemaster/deepseek_mtp_main_random_bf16"
     with VllmRunner(
             model,
             max_model_len=1024,
@@ -78,15 +72,8 @@ def test_pcp_dcp_mtp3_eager():
         runner.generate_greedy(prompts, 32)
 
 
-@pytest.mark.skip(
-    reason="vLLM PR-32118 break this",
-)
+@wait_until_npu_memory_free()
 def test_pcp_dcp_mtp3_piecewise_graph():
-    prompts = [
-        "The capital of France is", "Hello, my name is Tom, I am",
-        "The president of United States is", "AI future is"
-    ]
-    model = "wemaster/deepseek_mtp_main_random_bf16"
     with VllmRunner(
             model,
             max_model_len=1024,
@@ -109,15 +96,8 @@ def test_pcp_dcp_mtp3_piecewise_graph():
         runner.generate_greedy(prompts, 32)
 
 
-@pytest.mark.skip(
-    reason="vLLM PR-32118 break this",
-)
+@wait_until_npu_memory_free()
 def test_pcp_dcp_mtp3_full_graph():
-    prompts = [
-        "The capital of France is", "Hello, my name is Tom, I am",
-        "The president of United States is", "AI future is"
-    ]
-    model = "wemaster/deepseek_mtp_main_random_bf16"
     with VllmRunner(
             model,
             max_model_len=1024,
@@ -140,12 +120,8 @@ def test_pcp_dcp_mtp3_full_graph():
         runner.generate_greedy(prompts, 32)
 
 
+@wait_until_npu_memory_free()
 def test_dcp_mtp3_full_graph():
-    prompts = [
-        "The capital of France is", "Hello, my name is Tom, I am",
-        "The president of United States is", "AI future is"
-    ]
-    model = "wemaster/deepseek_mtp_main_random_bf16"
     with VllmRunner(
             model,
             max_model_len=1024,