failover: add method to manually reconnect

When all servers are down, a retry only happens after the retry
timeout. We want to be able to manually trigger reconnecting earlier
than that.

When the BitBoxApp iOS app goes into background, it kills TCP
connections shortly after. We want to be able to trigger a reconnecct
as soon as the iOS app comes to the foreground again.

failover/failover.go

@@ -121,6 +121,8 @@ type Failover[C Client] struct {
 	enableRetry   bool
 	subscriptions []func(client C, currentClientCounter int)
 
+	manualReconnect chan struct{}
+
 	closed   bool
 	closedMu sync.RWMutex
 }
@@ -145,6 +147,7 @@ func New[C Client](opts *Options[C]) *Failover[C] {
 		opts:               opts,
 		startServerIndex:   startServerIndex,
 		currentServerIndex: startServerIndex,
+		manualReconnect:    make(chan struct{}),
 	}
 }
 
@@ -157,7 +160,17 @@ func (f *Failover[C]) establishConnection() error {
 			if f.opts.OnRetry != nil {
 				go f.opts.OnRetry(f.lastErr)
 			}
-			time.Sleep(retryTimeout)
+
+			// Drain the manualReconnect channel to avoid stale reconnect signals
+			select {
+			case <-f.manualReconnect:
+			default:
+			}
+			// Wait for retry timeout or manual reconnect
+			select {
+			case <-time.After(retryTimeout):
+			case <-f.manualReconnect:
+			}
 		}
 		f.enableRetry = true
 
@@ -370,6 +383,16 @@ func (f *Failover[C]) isClosed() bool {
 	return f.closed
 }
 
+// ManualReconnect triggers a manual reconnect, non-blocking.
+// This re-tries connecting immediately without waiting for the retry timeout.
+// We we are not currently disconnected, this is a no-op.
+func (f *Failover[C]) ManualReconnect() {
+	select {
+	case f.manualReconnect <- struct{}{}:
+	default:
+	}
+}
+
 // Close closes the failover client and closes the current client, resulting in `ErrClosed` in all
 // future `Call` and `Subscribe` calls. It also calls `Close()` on the currently active client if
 // one exists.