[refactor] Adopt get_parallel() everywhere and close out the parallel wrapper surface (#30492)

test/registered/unit/model_executor/test_pool_configurator.py

@@ -10,6 +10,7 @@ import unittest
 from types import SimpleNamespace
 from unittest.mock import MagicMock, patch
 
+from sglang.srt.runtime_context import get_parallel
 from sglang.test.ci.ci_register import register_cpu_ci
 
 register_cpu_ci(est_time=10, suite="base-a-test-cpu")
@@ -27,10 +28,7 @@ def mock_cpu_env(kv_size=2, tp_size=1, swa_eviction_interval=4):
 
     with (
         patch("torch._utils._element_size", return_value=kv_size),
-        patch(
-            "sglang.srt.model_executor.pool_configurator.get_attention_tp_size",
-            return_value=tp_size,
-        ),
+        get_parallel().override(attn_tp_size=tp_size),
         envs.SGLANG_SWA_EVICTION_INTERVAL.override(swa_eviction_interval),
     ):
         yield