[scheduler] Feed watchdog after each scheduled item, drop top-of-loop feed

The main loop used to feed the watchdog unconditionally right after
Scheduler::call() returned, regardless of whether the scheduler had any
actual work to do. On an idle device this meant every outer loop
iteration paid the inline rate-limit check (load + sub + branch) for no
benefit.

Move the feed into Scheduler::execute_item_() so it fires only after a
scheduled callback actually runs, and covers both the main heap path
and the defer queue path (both go through execute_item_). This also
bounds the max feed gap during a burst of back-to-back scheduled items
by max(item_runtime) instead of sum(item_runtime).

The top-of-loop feed in Application::before_loop_tasks_() is now
unnecessary — when Scheduler::call does no work, the only elapsed time
is the sleep wake plus a few instructions, and when it does have work,
it fed the wdt as it went.

esphome/core/application.h

@@ -829,12 +829,13 @@ inline void ESPHOME_ALWAYS_INLINE Application::before_loop_tasks_(uint32_t loop_
   this->drain_wake_notifications_();
 #endif
 
-  // Process scheduled tasks
+  // Process scheduled tasks. Scheduler::call now feeds the watchdog itself
+  // after each scheduled item that actually runs, so we no longer need an
+  // unconditional feed here — when Scheduler::call has no work to do, the
+  // only elapsed time is a sleep wake + a few instructions, and when it does
+  // have work, it fed the wdt as it went.
   this->scheduler.call(loop_start_time);
 
-  // Feed the watchdog timer
-  this->feed_wdt(loop_start_time);
-
   // Process any pending enable_loop requests from ISRs
   // This must be done before marking in_loop_ = true to avoid race conditions
   if (this->has_pending_enable_loop_requests_) {

esphome/core/scheduler.cpp

@@ -739,7 +739,13 @@ uint32_t HOT Scheduler::execute_item_(SchedulerItem *item, uint32_t now) {
   App.set_current_component(item->component);
   WarnIfComponentBlockingGuard guard{item->component, now};
   item->callback();
-  return guard.finish();
+  uint32_t end = guard.finish();
+  // Feed the watchdog after each scheduled item (both main heap and defer
+  // queue paths go through here). A run of back-to-back callbacks cannot
+  // starve the wdt. The inline fast path is a load + sub + branch — nearly
+  // free when the 3 ms rate limit hasn't elapsed.
+  App.feed_wdt(end);
+  return end;
 }
 
 // Common implementation for cancel operations - handles locking