Support for parallel filter instances

[oplehto]

Right now the filter is a bit of a SPOF, although any disruption should be fairly minor due to auto-recovery mechanisms in the service management level (supervisord, k8s, systemd) and NATS buffering. At some point we will also start hitting the limits of how far we can scale the filter within the confines of a single physical server which is a slightly bigger concern.

To make things more scalable and redundant, we could run multiple filter instances in parallel and have the incoming data distributed across them. In practice I think it may be as simple as using QueueSubscribe (https://github.com/nats-io/go-nats#queue-groups) instead of Subscribe to subscribe to subjects in the filter.