Fix: Remove tensor pointers from autotune keys in 08-grouped-gemm tutorial

[niyunsheng]

New contributor declaration

• I am not making a trivial change, such as fixing a typo in a comment.

• I have written a PR description following these
  rules.

• I have run pre-commit run --from-ref origin/main --to-ref HEAD.

• Select one of the following.

  • I have added tests.
    • /test for lit tests
    • /unittest for C++ tests
    • /python/test for end-to-end tests
  • This PR does not need a test because FILL THIS IN.

• Select one of the following.

  • I have not added any lit tests.
  • The lit tests I have added follow these best practices,
    including the "tests should be minimal" section. (Usually running Python code
    and using the instructions it generates is not minimal.)

Description: This PR removes tensor objects (group_a_ptrs, group_b_ptrs, group_c_ptrs) from the @triton.autotune keys in the Grouped GEMM tutorial.

Why this matters:

1. Prevents Misleading Practices: Tutorials serve as reference templates. Passing tensors to autotune keys is a known anti-pattern. Fixing this prevents developers from propagating this flawed logic into production environments.
2. Fixes Overhead & Memory Leaks: Using tensors as keys relies on hashing their Python object IDs. Since these IDs change across dispatches, it triggers continuous re-benchmarking (100% cache miss) and causes VRAM leaks because the autotuner cache retains strong references to the temporary tensors.

Only the scalar group_size is retained as a valid cache key.