Improve network to connect up to 50 nodes

[ch1bo]

What & Why

Currently our network layer fails in connecting a large (> 10) number of nodes, at least in our local load tests. Within this feature we want to improve the hydra networking layer to be able to connect at least 50 nodes into a (still fully-connected) network to broadcast Hydra off-chain messages.

Requirements

• The local hydra-cluster can spin up and connect 50 hydra-nodes.
• The load test can be run with 50 nodes and does not time out.

[abailly-iohk]

A few issues currently prevent us from scaling the hydra-cluster benchmark to that number of Hydra nodes:

• Infrastructure: Running 50-100 nodes on a single machine does not really make sense and is not realistic. We need to be able to configure smaller VMs, each running a single Hydra node and associated Cardano node in BFT mode, and configure both networks. This should be done in an automated and configurable way in order to be able to run different scenarios easily, and requires some more configuration options for the CardanoNode and HydraNode executables
• Blockchain configuration: Current mainchain parameters for tx size and execution units are known to be insufficient for such a large number of participants. We should work on optimising our scripts' size and execution cost but in the meantime we should make sure network parameters are set to high enough values
• Script execution time: It seems that the actual execution time of the CollectCom transaction's scripts is too high, leading to exceedingly long validation time (in the order of minutes...). This is a show-stopper that needs to investigated and addressed.

[abailly-iohk]

Also, I think that solving point 1 from the previous comment will be useful for deploying demo/test Hydra networks.

[abailly-iohk]

We are mostly concerned right now with assessing the performance of the layer 2, so we might as well resurrect our MockChain component to handle the on-chain part. This would simplify deployment as we would only need a single instance of the mockchain component running, without impacting the validity of our observations about the performance of the layer 2.

[ch1bo]

So we solved the issue with ourobouros-network and we can run 50 nodes benchmarks by increasing the execution budgets and slot lenghts to accomodate the actual validation (of collectCom). We could also use the mock-chain, which we ended up not merging now though.