bugfix: fix the register overflow issue for topk renorm kernels on blackwell (#1597)

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

## 📌 Description

On blackwell, we will encounter `failed with error code too many
resources requested for launch` (register overflow) issue when using the
`TopKRenormProbKernel` (another issue that only happens on blackwell).
This PR fixes the issue by reducing the loop unroll number to reduce the
number of registers used in the kernel.

This PR is based upon #1596 , these two PRs together will fix the failed
UTs on blackwell.

## 🔍 Related Issues

<!-- Link any related issues here -->

## 🚀 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

- [x] I have installed `pre-commit` by running `pip install pre-commit`
(or used your preferred method).
- [x] I have installed the hooks with `pre-commit install`.
- [x] 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

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

## Reviewer Notes

cc @bkryu

Author: yzh119

include/flashinfer/sampling.cuh

@@ -1891,7 +1891,7 @@ __global__ void TopKRenormProbKernel(DType* probs, DType* renormed_prob, IdType*
       ValueCount<float> aggregate_gt_pivot_0{0, 0}, aggregate_gt_pivot_1{0, 0};
       min_gt_low = high;
       max_le_high = low;
-#pragma unroll 2
+#pragma unroll 1
       for (uint32_t i = 0; i < ceil_div(d, BLOCK_THREADS * VEC_SIZE); ++i) {
         probs_vec.fill(0);
         if ((i * BLOCK_THREADS + tx) * VEC_SIZE < d) {