try to avoid bloat

Replace the temporary std::vector copy with in-place compaction using a
read/write pointer pattern. This avoids a heap allocation+deallocation
cycle during scheduler cleanup, reducing heap fragmentation on
long-running ESP devices.

The new approach compacts valid items forward in the existing vector,
recycles removed items as they are encountered, then resizes the vector
(no reallocation since size only shrinks). Same O(n) complexity, same
behavior, zero allocations.

esphome/core/scheduler.cpp

@@ -390,20 +390,19 @@ void Scheduler::full_cleanup_removed_items_() {
   // 4. No operations inside can block or take other locks, so no deadlock risk
   LockGuard guard{this->lock_};
 
-  std::vector<std::unique_ptr<SchedulerItem>> valid_items;
-
-  // Move all non-removed items to valid_items, recycle removed ones
-  for (auto &item : this->items_) {
-    if (!is_item_removed_(item.get())) {
-      valid_items.push_back(std::move(item));
+  // Compact in-place: move valid items forward, recycle removed ones
+  size_t write = 0;
+  for (size_t read = 0; read < this->items_.size(); ++read) {
+    if (!is_item_removed_(this->items_[read].get())) {
+      if (write != read) {
+        this->items_[write] = std::move(this->items_[read]);
+      }
+      ++write;
     } else {
-      // Recycle removed items
-      this->recycle_item_main_loop_(std::move(item));
+      this->recycle_item_main_loop_(std::move(this->items_[read]));
     }
   }
-
-  // Replace items_ with the filtered list
-  this->items_ = std::move(valid_items);
+  this->items_.erase(this->items_.begin() + write, this->items_.end());
   // Rebuild the heap structure since items are no longer in heap order
   std::make_heap(this->items_.begin(), this->items_.end(), SchedulerItem::cmp);
   this->to_remove_ = 0;