[TRTLLM-10030][chore] promote SampleState to TypeVar + typing fixes (NVIDIA#11281)

Signed-off-by: ixlmar <206748156+ixlmar@users.noreply.github.com>

Author: ixlmar

tensorrt_llm/_torch/pyexecutor/llm_request.py

@@ -1,10 +1,14 @@
 from copy import copy, deepcopy
 from dataclasses import dataclass
-from typing import Any, Dict, List, Optional, Union
+from typing import TYPE_CHECKING, Any, Dict, List, Optional, Union
 
 import torch
 
 import tensorrt_llm.bindings
+
+if TYPE_CHECKING:
+    from tensorrt_llm._torch.pyexecutor.sampler import Strategy
+
 from tensorrt_llm._torch.shared_tensor import SharedTensorContainer
 from tensorrt_llm.bindings import executor as tllm_executor
 from tensorrt_llm.executor.result import TokenLogprobs
@@ -583,6 +587,8 @@ def __init__(
             additional_outputs=additional_outputs)
         self.child_requests = []
 
+        self._py_sampling_strategy: "Strategy | None" = None
+
         self._py_embedding_bias_1d: Optional[torch.Tensor] = None
         if hasattr(self, 'embedding_bias') and self.embedding_bias is not None:
             # Pre-squeeze to 1D if needed (remove batch dimension)