[FEAT] - Drop unused consensus protocols and code

[ch1bo]

A discussion on the recent Tachys CIP reminded me of my recent question in one of the consensus working groups:

Why is there still block production code for old block formats / eras?

See this issue as a discussion (the github discussions seem to be stale on this repo) on what parts of ouroboros-consensus are actually unused by the Cardano mainnet and hence could be dropped?

So far I heard:

• Block production of Byron blocks
• Ouroboros BFT was never used on mainnet

[dcoutts]

It is currently not possible for a node in a mainnet configuration to start in the latest era. Thus currently, new partner chains must start in Byron and hard fork through all the eras. This requires block production.

Of course this could and should be improved so that partner chains can start in the latest (or perhaps previous) era in a mainnet configuration. My colleagues and I are interested in contributing improvements to make this possible, so that it becomes easier for anyone in the community to start new partner chains.

In the meantime, I would ask that we hold off on removing block production support. But I do agree that ultimately it's a reasonable cleanup.

I think it is also a useful data point to see how many changes the old protocols (and BFT) have seen in the last few years, as a measure of the maintenance cost. For example, a quick look at the git history for Ouroboros/Consensus/Protocol/BFT.hs shows that it has had no non-trivial changes in the last 3 years. All the changes are things like reformatting, language pragmas, from patches that touch minor things widely across the tree.

Historically the reason we kept the BFT module was:

1. to keep ourselves honest about the interfaces by having at least one example of doing things differently
2. the hope that someone would listen to us and re-use the code base (as we had designed it to be reused) for some non-mainnet use case, like private enterprise chains. This was the same reason we demoed a pokemon ledger.