[muh] add tuning_unique_by_key.cuh: BI-V100 tuning for unique_by_key

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

muh/include/muh/tuning/tuning_unique_by_key.cuh

@@ -0,0 +1,52 @@
+// muh/include/muh/tuning/tuning_unique_by_key.cuh — BI-V100
+//
+// Mirrors: cccl_upstream/cub/cub/device/dispatch/tuning/tuning_unique_by_key.cuh
+// CCCL SM100: 51 specializations, 7 SMEM overflow (max tile=57344)
+//
+// vllm relevance: deduplicated token sequences
+
+#pragma once
+
+#include "muh/hardware.cuh"
+#include "muh/tuning/common.cuh"
+
+namespace muh::tuning::unique_by_key {
+
+struct UniqueByKeyPolicy {
+  int threads_per_block;
+  int items_per_thread;
+  BlockLoadAlgorithm load_algorithm;
+  CacheLoadModifier load_modifier;
+  BlockScanAlgorithm scan_algorithm;
+  LookbackDelayPolicy delay;
+};
+
+struct policy_selector {
+  int key_size;
+  int value_size;
+  int offset_size;
+
+  constexpr UniqueByKeyPolicy operator()(const hardware_capability& hw) const {
+    int pair_size = key_size + value_size;
+    
+    int threads = 256;
+    int items = 12;
+    
+    if (pair_size <= 4) {
+      threads = 320; items = 16;
+    } else if (pair_size <= 8) {
+      threads = 256; items = 12;
+    } else {
+      threads = 192; items = 8;
+    }
+    
+    while (threads * items * pair_size > hw.max_shared_memory_per_block && items > 1)
+      items--;
+
+    return {threads, items, BLOCK_LOAD_WARP_TRANSPOSE, LOAD_DEFAULT,
+            BLOCK_SCAN_WARP_SCANS,
+            {LookbackDelayAlgorithm::exponential_backon, 350, 450}};
+  }
+};
+
+} // namespace muh::tuning::unique_by_key