Make deliveries refused while half-open wait a cooldown (closes #306)

While the breaker was half-open, Allow refused every delivery but the
probe and CooldownRemaining returned zero, so each queued task for the
target went straight back onto the retry channel and rewrote its status
on every pass until the probe finished.

CooldownRemaining now returns the whole cooldown while half-open, so a
refused delivery waits that long. A refused delivery already at
retrying is not written again, so the retry counter now moves only
when a refusal moves a delivery into retrying.

Model: opus-5-5

internal/delivery/circuit_breaker.go

@@ -76,12 +76,20 @@ func (cb *CircuitBreaker) Allow() bool {
 	}
 }
 
-// CooldownRemaining returns how much time is left before
-// an open circuit transitions to half-open.
+// CooldownRemaining returns how long a delivery that Allow refused
+// should wait before it is tried again. Closed, it returns zero.
+// Open, it returns what is left of the cooldown, or zero once that
+// has passed. Half-open, it returns the whole cooldown: the one
+// probe delivery is still in flight, and if it fails the circuit
+// reopens for that long.
 func (cb *CircuitBreaker) CooldownRemaining() time.Duration {
 	cb.mu.Lock()
 	defer cb.mu.Unlock()
 
+	if cb.state == CircuitHalfOpen {
+		return cb.cooldown
+	}
+
 	if cb.state != CircuitOpen {
 		return 0
 	}