generalizing attsim

[raulk]

attsim's initial goal was to simulate the network-wide improvements we'd expect to see on mainnet after optimizing Prysm's validation pipeline: OffchainLabs/prysm#15882

That said, we always had an eye on generalizing attsim so it can serve as a basis for simulating new attestation broadcast protocols and topologies.

In my view, this second goal requires, at least:

1. Simulating entire networks, not just a single subnet.
2. Generalizing the topology formation, so that we can arrange nodes under different configurations (e.g. number of sunets, number of subscriptions per node, etc.)
3. Parameterizing the publish, so that we can test different strategies and fanouts.
4. Some basic awareness of attestation semantics, e.g. to simulate attestation aggreation at origin

Dropping a whiteboard with some dirty sketches -- not super informational in itself, but useful as a future reference for discussions