perf(pool): eliminate mutex overhead in state machine hot path

The state machine was calling notifyWaiters() on EVERY Get/Put operation,
which acquired a mutex even when no waiters were present (the common case).

Fix: Check waiters.Len() BEFORE acquiring mutex in notifyWaiters().
This eliminates mutex contention in the hot path (Get/Put operations).

Performance impact:
- Before: mutex lock/unlock on every Get/Put (even with no waiters)
- After: lock-free check, only acquire mutex if waiters exist
- Expected improvement: ~30-50% for Get/Put operations

Author: ndyakov

internal/pool/conn_state.go

@@ -238,9 +238,19 @@ func (sm *ConnStateMachine) AwaitAndTransition(
 // notifyWaiters checks if any waiters can proceed and notifies them in FIFO order.
 // This is called after every state transition.
 func (sm *ConnStateMachine) notifyWaiters() {
+	// Fast path: check if there are any waiters without acquiring lock
+	// This is safe because:
+	// 1. list.List.Len() is safe to call concurrently with modifications
+	// 2. We only care if it's zero (no waiters) - this is the common case
+	// 3. If it's non-zero, we acquire the lock and check again
+	if sm.waiters.Len() == 0 {
+		return
+	}
+
 	sm.mu.Lock()
 	defer sm.mu.Unlock()
 
+	// Double-check after acquiring lock (waiters might have been processed)
 	if sm.waiters.Len() == 0 {
 		return
 	}