[muh] add tuning_segmented_sort.cuh: BI-V100 tuning for segmented_sort

Translated from CCCL with SMEM overflow protection.
All SM100 values checked against 48KB limit.

muh/include/muh/tuning/tuning_segmented_sort.cuh

@@ -0,0 +1,46 @@
+// muh/include/muh/tuning/tuning_segmented_sort.cuh — BI-V100
+//
+// Mirrors: cccl_upstream/cub/cub/device/dispatch/tuning/tuning_segmented_sort.cuh
+// CCCL: chained_policy with per-segment-size dispatch (large/medium/small)
+//
+// vllm relevance: per-sequence token ranking
+
+#pragma once
+
+#include "muh/hardware.cuh"
+#include "muh/tuning/common.cuh"
+
+namespace muh::tuning::segmented_sort {
+
+struct SegmentedSortPolicy {
+  int large_threads;
+  int large_items;
+  int medium_threads;
+  int medium_items;
+  int small_threads;
+  int small_items;
+  BlockLoadAlgorithm load_algorithm;
+};
+
+struct policy_selector {
+  int key_size;
+  int value_size;
+  bool keys_only;
+
+  constexpr SegmentedSortPolicy operator()(const hardware_capability& hw) const {
+    int pair_size = key_size + (keys_only ? 0 : value_size);
+    
+    // Large segments: full block sort
+    auto [lg_items, lg_threads] = scale_mem_bound(256, 11, pair_size);
+    // Medium: partial
+    auto [md_items, md_threads] = scale_mem_bound(128, 15, pair_size);
+    // Small: warp sort
+    int sm_threads = 32;
+    int sm_items = 4;
+
+    return {lg_threads, lg_items, md_threads, md_items, 
+            sm_threads, sm_items, BLOCK_LOAD_WARP_TRANSPOSE};
+  }
+};
+
+} // namespace muh::tuning::segmented_sort