vllm

[Docs] Update `CacheConfig` block_size docstring to remove inaccurate limit when using CUDA

#35632

Merged