[Bugfix] Change max_size for LRU (#1349)

<!-- .github/pull_request_template.md -->

## 📌 Description

LRU cache eviction blocks frameworks for cuda graph replay. Setting to a
higher size.

```Fatal Python error: Segmentation fault

Thread 0x00007f62e7ffe6c0 (most recent call first):
  File "/opt/mycode/python/sglang/srt/managers/scheduler.py", line 2215 in watchdog_thread
  File "/usr/lib/python3.12/threading.py", line 1010 in run
  File "/usr/lib/python3.12/threading.py", line 1073 in _bootstrap_inner
  File "/usr/lib/python3.12/threading.py", line 1030 in _bootstrap

Current thread 0x00007f62ebfff6c0 (most recent call first):
  File "/usr/local/lib/python3.12/dist-packages/torch/cuda/graphs.py", line 88 in replay
  File "/opt/mycode/python/sglang/srt/model_executor/cuda_graph_runner.py", line 752 in replay
  File "/opt/mycode/python/sglang/srt/model_executor/model_runner.py", line 1633 in _forward_raw
  File "/opt/mycode/python/sglang/srt/model_executor/model_runner.py", line 1606 in forward
  File "/opt/mycode/python/sglang/srt/managers/tp_worker.py", line 234 in forward_batch_generation
  File "/opt/mycode/python/sglang/srt/managers/tp_worker_overlap_thread.py", line 175[2025-07-29 07:12:06] INFO:     127.0.0.1:49294 - "POST /generate HTTP/1.1" 200 OK
 in forward_thread_func_
  File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 116 in decorate_context
  File "/opt/mycode/python/sglang/srt/managers/tp_worker_overlap_thread.py", line 140 in forward_thread_func
  File "/usr/lib/python3.12/threading.py", line 1010 in run
  File "/usr/lib/python3.12/threading.py", line 1073 in _bootstrap_inner
  File "/usr/lib/python3.12/threading.py", line 1030 in _bootstrap

Thread 0x00007f72fffe46c0 (most recent call first):
  File "/usr/lib/python3.12/threading.py", line 359 in wait
  File Fatal Python error: "Segmentation fault/usr/lib/python3.12/threading.py

"Fatal Python error: Thread 0x, line Segmentation fault00007f8375ffd6c0655

```


## 🔍 Related Issues


## 🚀 Pull Request Checklist

Thank you for contributing to FlashInfer! Before we review your pull
request, please make sure the following items are complete.

### ✅ Pre-commit Checks

- [ ] I have installed `pre-commit` by running `pip install pre-commit`
(or used your preferred method).
- [ ] I have installed the hooks with `pre-commit install`.
- [ ] I have run the hooks manually with `pre-commit run --all-files`
and fixed any reported issues.

> If you are unsure about how to set up `pre-commit`, see [the
pre-commit documentation](https://pre-commit.com/).

## 🧪 Tests

- [ ] Tests have been added or updated as needed.
- [ ] All tests are passing (`unittest`, etc.).

## Reviewer Notes

<!-- Optional: anything you'd like reviewers to focus on, concerns, etc.
-->

Author: elfiegg

flashinfer/gemm.py

@@ -834,7 +834,7 @@ def _validate_fp8_output_dtype(dtype: torch.dtype):
         )
 
 
-@functools.lru_cache(maxsize=128)
+@functools.cache
 def build_cudnn_gemm_block_scale_dequantize_graph(
     a_shape,
     a_stride,
@@ -946,7 +946,7 @@ def execute_cudnn_gemm_fp4_graph(graph, a, b, a_descale, b_descale, alpha, c_fin
     graph.execute(variant_pack, workspace, handle=_get_cudnn_handle(stream))
 
 
-@functools.lru_cache(maxsize=128)
+@functools.cache
 def build_cudnn_gemm_with_per_tensor_q_graph(
     a_shape, a_stride, b_shape, b_stride, a_type, b_type, o_type, device
 ):