Support experiments on Lily

[iand]

This is a placeholder to capture thoughts about testing Lily. The principle is the same but the details differ from testing gateways. We need to run various instances, apply load and then capture measurements. For Lily the load would be starting tasks (which can be done via API).

Could help us answer questions about resources needed to run lily: cpu/disk io/network io/memory. Also could compare releases to identify performance regressions. Potentially could compare algorithm improvements on known sections of the chain (such as miner spam era)