Add Page Box layer for C7 class optimization

- Implement tiny_page_box.c/h: per-thread page cache between UC and Shared Pool
- Integrate Page Box into Unified Cache refill path
- Remove legacy SuperSlab implementation (merged into smallmid)
- Add HAKMEM_TINY_PAGE_BOX_CLASSES env var for selective class enabling
- Update bench_random_mixed.c with Page Box statistics

Current status: Implementation safe, no regressions.
Page Box ON/OFF shows minimal difference - pool strategy needs tuning.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

core/front/tiny_unified_cache.h

@@ -40,6 +40,11 @@ extern _Atomic uint64_t g_unified_cache_hits_global;
 extern _Atomic uint64_t g_unified_cache_misses_global;
 extern _Atomic uint64_t g_unified_cache_refill_cycles_global;
 
+// Per-class counters（観測用 Box、ENV でのみ有効）
+extern _Atomic uint64_t g_unified_cache_hits_by_class[TINY_NUM_CLASSES];
+extern _Atomic uint64_t g_unified_cache_misses_by_class[TINY_NUM_CLASSES];
+extern _Atomic uint64_t g_unified_cache_refill_cycles_by_class[TINY_NUM_CLASSES];
+
 // Print statistics function
 void unified_cache_print_measurements(void);
 
@@ -102,10 +107,15 @@ static inline size_t unified_capacity(int class_idx) {
         snprintf(env_name, sizeof(env_name), "HAKMEM_TINY_UNIFIED_C%d", class_idx);
         const char* e = getenv(env_name);
 
-        // Default: Hot_2048 strategy (C2/C3=2048, others=64)
+        // Default: Hot_2048 strategy
+        //   - C2/C3 (128B/256B): 2048 slots（超ホット Tiny）
+        //   - C5/C6/C7 (>=129B): 128 slots（Mixed 用に拡張）
+        //   - その他: 64 slots（コールド）
         size_t default_cap = 64;  // Cold classes
         if (class_idx == 2 || class_idx == 3) {
-            default_cap = 2048;  // Hot classes (128B, 256B)
+            default_cap = 2048;  // Hot Tiny classes (128B, 256B)
+        } else if (class_idx >= 5 && class_idx <= 7) {
+            default_cap = 128;   // Mixed workload classes (C5-C7: 129B-1024B)
         }
 
         g_cap[class_idx] = (e && *e) ? (size_t)atoi(e) : default_cap;
@@ -265,9 +275,13 @@ static inline hak_base_ptr_t unified_cache_pop_or_refill(int class_idx) {
 #if !HAKMEM_BUILD_RELEASE
         g_unified_cache_hit[class_idx]++;
 #endif
-        // Performance measurement: count cache hits
+        // Performance measurement: count cache hits（ENV 有効時のみ）
         if (__builtin_expect(unified_cache_measure_check(), 0)) {
-            atomic_fetch_add_explicit(&g_unified_cache_hits_global, 1, memory_order_relaxed);
+            atomic_fetch_add_explicit(&g_unified_cache_hits_global,
+                                      1, memory_order_relaxed);
+            // Per-class ヒット（C5–C7 の利用率も可視化）
+            atomic_fetch_add_explicit(&g_unified_cache_hits_by_class[class_idx],
+                                      1, memory_order_relaxed);
         }
         return HAK_BASE_FROM_RAW(base);  // Hit! (2-3 cache misses total)
     }