chore|research(p2p): investigate Gossipsub behaviour and tune parameters

[rymnc]

Description

We currently use libp2p's GossipSub to disseminate transactions for the txpool. Transactions can vary between 1kb ~ 200kb with 1kb being the most common one. GossipSub is a protocol that has to be tuned according to the "mode" of the message size.

Refer to the following doc for some more details on how to conduct simulations/research on decision about gossipsub params:

• https://github.com/waku-org/research/blob/master/docs/message-propagation.md#theory
• we can reuse the results from https://github.com/waku-org/research/blob/master/docs/message-propagation.md#results but we should conduct our own analysis with data from metrics about average message size, average message frequency etc.

Definition of done

• Simulations ran in live nodes spread across geographic locations to obtain data about message propagation time, size of message etc
• tuning of gossipsub params and re-run simulation for the new values
• overall reduction in bandwidth consumed by gossipsub, with lower latency for tx propagation

Why?

our executor is pretty fast, eventually the bottleneck for our tps will be consumption of transactions via gossipsub so it is useful to reduce this bottleneck.