Support executing start multiple times from same instance of producer

[tiny-dancer]

Ran into an issue in lambda where every execution after the cold start would fail due to panic: close of closed channel. Therefore leading to a 50% failure rate.

Root cause was defining the producer as a global variable outside the Handler function. This pull request assumes Start() desires idempotency. Please correct me if that is not desired and the appropriate solution is to call .New for every execution within the Handler.

kinesisProducer = producer.New(&producer.Config{
		StreamName:    kinesisStream,
		Client:        kinesisClient,
		BatchCount:    500,
		FlushInterval: 60 * time.Second,
		Logger:        &kplogrus.Logger{Logger: logs}, 
	})

func Handler(ctx context.Context, sqsEvent events.SQSEvent) error {
	kinesisProducer.Start() // <-- always fail after initial cold start

	var wg sync.WaitGroup
	wg.Add(1)

	// concurrently handle failures
	go processPutFailures(&wg)

	processSqsEvent(sqsEvent)

	kinesisProducer.Stop()

	wg.Wait()
        ...
}