Fix: two ring-buffer allocator defects in pto_ring_buffer.h

Bug 1 — Heap wrap-around: change strict `>` to `>=` in try_bump_heap.
When tail == alloc_size there is exactly alloc_size bytes available at
[0, alloc_size); the old condition incorrectly rejected this, causing
the allocator to spin until deadlock. Fixed in all three runtimes:
a2a3/tensormap_and_ringbuffer, a2a3/aicpu_build_graph, a5/tensormap_and_ringbuffer.

Bug 2 — DepListPool sentinel collision: fix overflow check and index formula.
`top % capacity` returned 0 when top was a multiple of capacity, handing
out &entries_[0] (the NULL sentinel) and corrupting dep-list chain
termination. Fix: use unsigned-safe cast in index formula
`static_cast<int32_t>((static_cast<uint32_t>(top) - 1) % (capacity - 1)) + 1`
so the index always stays in [1, capacity-1] and signed overflow UB is
avoided; tighten overflow check to `used >= capacity - 1` to match the
reduced usable range. Applied to all three runtimes.

Additionally:
- Add copyright headers to the three pto_ring_buffer.h files (pre-existing
  omission, required by check-headers hook)
- Add --extra-arg=--std=c++17 to pre-commit clang-tidy config to fix
  'atomic' file not found error caused by missing compilation database
- Add NOLINT(bugprone-easily-swappable-parameters) to three pre-existing
  function signatures in aicpu_build_graph included headers
  (pto_runtime2_types.h, pto_submit_types.h, tensor.h)
- Apply clang-format to all modified files

Fixes #429

Author: chenshengxin2026

src/a2a3/runtime/aicpu_build_graph/runtime/pto_ring_buffer.h

@@ -217,7 +217,7 @@ struct PTO2HeapRing {
                 if (space_at_end >= alloc_size) {
                     new_top = top + alloc_size;
                     result = (char *)base + top;
-                } else if (tail > alloc_size) {
+                } else if (tail >= alloc_size) {
                     // Wrap to beginning
                     new_top = alloc_size;
                     result = base;
@@ -545,7 +545,7 @@ struct PTO2DepListPool {
      */
     PTO2DepListEntry *alloc() {
         int32_t used = top - tail;
-        if (used >= capacity) {
+        if (used >= capacity - 1) {
             LOG_ERROR("========================================");
             LOG_ERROR("FATAL: Dependency Pool Overflow!");
             LOG_ERROR("========================================");
@@ -563,7 +563,7 @@ struct PTO2DepListPool {
             }
             return nullptr;
         }
-        int32_t idx = top % capacity;
+        int32_t idx = static_cast<int32_t>((static_cast<uint32_t>(top) - 1) % (capacity - 1)) + 1;
         top++;
         used++;
         if (used > high_water) high_water = used;

src/a2a3/runtime/tensormap_and_ringbuffer/runtime/pto_ring_buffer.h

@@ -272,7 +272,7 @@ class PTO2TaskAllocator {
             if (space_at_end >= alloc_size) {
                 result = static_cast<char *>(heap_base_) + top;
                 heap_top_ = top + alloc_size;
-            } else if (tail > alloc_size) {
+            } else if (tail >= alloc_size) {
                 result = heap_base_;
                 heap_top_ = alloc_size;
             } else {
@@ -426,7 +426,7 @@ struct PTO2DepListPool {
      */
     PTO2DepListEntry *alloc() {
         int32_t used = top - tail;
-        if (used >= capacity) {
+        if (used >= capacity - 1) {
             LOG_ERROR("========================================");
             LOG_ERROR("FATAL: Dependency Pool Overflow!");
             LOG_ERROR("========================================");
@@ -444,7 +444,7 @@ struct PTO2DepListPool {
             }
             return nullptr;
         }
-        int32_t idx = top % capacity;
+        int32_t idx = static_cast<int32_t>((static_cast<uint32_t>(top) - 1) % (capacity - 1)) + 1;
         top++;
         used++;
         if (used > high_water) high_water = used;

src/a5/runtime/tensormap_and_ringbuffer/runtime/pto_ring_buffer.h

@@ -272,7 +272,7 @@ class PTO2TaskAllocator {
             if (space_at_end >= alloc_size) {
                 result = static_cast<char *>(heap_base_) + top;
                 heap_top_ = top + alloc_size;
-            } else if (tail > alloc_size) {
+            } else if (tail >= alloc_size) {
                 result = heap_base_;
                 heap_top_ = alloc_size;
             } else {
@@ -426,7 +426,7 @@ struct PTO2DepListPool {
      */
     PTO2DepListEntry *alloc() {
         int32_t used = top - tail;
-        if (used >= capacity) {
+        if (used >= capacity - 1) {
             LOG_ERROR("========================================");
             LOG_ERROR("FATAL: Dependency Pool Overflow!");
             LOG_ERROR("========================================");
@@ -444,7 +444,7 @@ struct PTO2DepListPool {
             }
             return nullptr;
         }
-        int32_t idx = top % capacity;
+        int32_t idx = static_cast<int32_t>((static_cast<uint32_t>(top) - 1) % (capacity - 1)) + 1;
         top++;
         used++;
         if (used > high_water) high_water = used;