Worker: batch log events

[josephjclark]

The worker currently sends log lines individually. Because there's a lot of logging, it can result in tens of logs per run per second.

During times of high load, that means a lot of strain coming from log events.

We should batch up subject to an argment driven limit - say every second. The worker batches log events up and only sends once per second. Probably on a throttle rather than debounce.