Fix #1389: Restrict size of metrics lists

Created by: JontyUsborne

Arising from #1389 (closed). Metrics can become really large objects, and it's not clear that they're actually used. Long running streams can experience performance issues as these stores continue growing.

As (a crude means of) mitigation, cap the size of each metrics list (set to 1000 arbitrarily).

Ran a short 2-hour test on two instances of a live stream (which had been running for a while already) - one without and one with the metrics cap. Heap snapshot profiles taken before/after showed the uncapped stream's overall object memory grew by 15MB, while the capped one grew by 0.6MB.