Create a more performant tracer for a network wide trace

[evan-forbes]

Tracing the entire network would likely greatly speed up debugging certain things about gossiping because we would be able to examine the lifecycle of each blockpart, tx, and vote.

When we attempt to trace the entire network with a single influx instance, we cause the network's throughput to decrease substantially. This makes the data collected unreliable. It might be possible to fix the issue by creating multiple influx instances and/or increases the resources of each node, however spinning up multiple influx instances is cumbersome, and we're already pushinght limits of resources that can be alloacted in testground.

Therefore, we should create a performant local only tracer option. This option could work similarly to the current option, but intead of writing to an external database, we would write files locally and then serve them / push them to a server some point after the trace period or the experiment concludes.