Bugfix: some typos in Persistent kernel (#1562)

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

## 📌 Description

tests pass 
## 🔍 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

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

Author: Edenzzzz

include/flashinfer/attention/persistent.cuh

@@ -266,7 +266,7 @@ struct BlockBatchPagedAttentionPersistent {
       const auto [q_indptr, kv_indptr, o_indptr, q_len, kv_len, packed_qo_start, kv_start, kv_end,
                   kv_head_idx, len_kv_chunk] = get_block_coord(params, work_idx);
 
-      const uint32_t kv_chunk_idx = ceil_div(kv_start, len_kv_chunk);
+      const uint32_t kv_chunk_idx = kv_start / len_kv_chunk;
       const uint32_t num_kv_chunks = ceil_div(
           CAUSAL
               ? min((kv_len - q_len) + (packed_qo_start + cluster_tile_q) / gqa_group_size, kv_len)

include/flashinfer/attention/scheduler.cuh

@@ -1189,8 +1189,7 @@ inline cudaError_t TwoStageHolisticPlan(void* float_buffer, size_t float_workspa
         cluster_kv_start(num_clusters, std::vector<IdType>()),
         cluster_kv_end(num_clusters, std::vector<IdType>()),
         cluster_kv_head_idx(num_clusters, std::vector<IdType>()),
-        cluster_partial_indptr(num_clusters, std::vector<IdType>()),
-        cluster_len_kv_chunk(num_clusters, std::vector<IdType>());
+        cluster_partial_indptr(num_clusters, std::vector<IdType>());
 
     for (auto& [i, qo_len, kv_len] : idx_qo_kv_len_vec[task]) {
       int packed_qo_len = qo_len * gqa_group_size;
@@ -1218,7 +1217,6 @@ inline cudaError_t TwoStageHolisticPlan(void* float_buffer, size_t float_workspa
             cluster_kv_indptr[cluster_idx].push_back(kv_indptr_h[i]);
 
             // use kv_chunk to rematerize num_kv_tiles and kv_tile_idx
-            cluster_len_kv_chunk[cluster_idx].push_back(kv_len_limit);
             cluster_partial_indptr[cluster_idx].push_back(partial_o_nnz);
 
             cluster_q_start[cluster_idx].push_back(qo_tile_idx * cluster_tile_q);
@@ -1265,7 +1263,6 @@ inline cudaError_t TwoStageHolisticPlan(void* float_buffer, size_t float_workspa
     auto kv_start_vec = flatten(cluster_kv_start, total_num_works);
     auto kv_end_vec = flatten(cluster_kv_end, total_num_works);
     auto kv_head_idx_vec = flatten(cluster_kv_head_idx, total_num_works);
-    auto len_kv_chunk_vec = flatten(cluster_len_kv_chunk, total_num_works);
 
     plan_info.tasks[task].q_indptr_offset =
         int_allocator.aligned_alloc_offset(sizeof(IdType) * max_total_num_works, 16, "q_indptr");