Add vLLM v0.18.1 source tree with KV transfer abort fix

third_party/vllm/ now tracked in git for direct patch management.
Based on vLLM v0.18.1 release with one patch applied:

  vllm/v1/core/sched/scheduler.py:
    Replace fatal assert with graceful skip when KV transfer callback
    arrives for an already-aborted request during PD disaggregated serving.

Future vLLM modifications should be made directly in third_party/vllm/
and committed normally. The patches/ directory is kept as documentation
of what changed from upstream.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

third_party/vllm/tests/test_seed_behavior.py

@@ -0,0 +1,25 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import random
+
+import numpy as np
+import torch
+
+from vllm.platforms.interface import Platform
+
+
+def test_seed_behavior():
+    # Test with a specific seed
+    Platform.seed_everything(42)
+    random_value_1 = random.randint(0, 100)
+    np_random_value_1 = np.random.randint(0, 100)
+    torch_random_value_1 = torch.randint(0, 100, (1,)).item()
+
+    Platform.seed_everything(42)
+    random_value_2 = random.randint(0, 100)
+    np_random_value_2 = np.random.randint(0, 100)
+    torch_random_value_2 = torch.randint(0, 100, (1,)).item()
+
+    assert random_value_1 == random_value_2
+    assert np_random_value_1 == np_random_value_2
+    assert torch_random_value_1 == torch_random_value_2