ACK / NACK semantics to make queues truly durable

[kshitij-g]

As per docs (https://rivet.dev/docs/actors/queues/), messages are removed from queue storage when they are received, not when they are completed. This poses a problem when the actor crashes immediately after receiving an event and crashing before it could complete the next step inside a workflow. complete() provides request/reply semantics but truly durability and replayability need ack/nack otherwise durability claim of using workflows is also incorrect.

[linear]

RVT-6342

[kshitij-g]

I guess workflows can be used to make them durable, but having ack / nack will provide additional flexibility.

In case and actor wants to skip a few queue messages and retain them for later processing, i e. not necessarily do sequential processing, then the ack / nack will be very helpful.