bench: add constant rate mode to test failover

[costela]

Proposed change

Add a "constant rate" mode for the bench subcommand.

Use case

The current benchmark "as fast as it gets" mode is great, but it doesn't cover all cases one might want to inspect.

One such case is testing fail-over behavior of a nats setup. This could be better tested if we had - e.g. - a "constant rate" mode, that just constantly writes messages (for --pub) and counts them on the other side (for --pub).

This way one can fire messages at different rates and inspect either drop-rate or queue-fill-rate (depending on whether one is using nats streaming or not).

In our specific case we have leaves running on bare-metal and publishing (without streams) to a cluster on k8s. Cluster rollouts currently cause leaves to disconnect/reconnect and drop a few messages in between.

Contribution

I'd love to give it a try!