Spike: How to measure the delay between a node accepting a message and the message becoming available on other nodes

[api-Hypernova]

Description:

The primary purpose here is the core component to a system which can detect bad node operators. This is not intended for benchmarking message replication.

As such, we've decided that time-to-visibility can be the reference metric. This simplifies things signifiantly from a system where we need to perform duplicate local/remote reads/writes to cancel each other out and isolate the replication timing,

I have a test for this already written up as part of #1377, but execution is currently blocked due to Node SDK-level bugs. I'm working on getting visibility on those bugs and once I have an ETA on resolution we can decide whether to go with that or use an xdbg-based solution (which would work perfectly well).