Cardano Updates

2496: buildkite: fix exit code in nightly ThreadNet script r=nfrisby a=nfrisby

:face_palm:

This BuildKite demonstrates the fix works. https://buildkite.com/input-output-hk/ouroboros-network-nightly/builds/186#037a7df9-04df-4f44-a6eb-2db0af2e6e5a (It had an additional commit that reduced the nightly invocations to just `RealTPraos` and reduced `RealTPraos` to just a coin toss.)

Co-authored-by: Nicolas Frisby <[email protected]>

2493: Use commads for db-validator r=nc6 a=kderme

https://github.com/input-output-hk/ouroboros-network/issues/2460

~This probably breaks a nightly test.~ fixed

An example to use it with mainnet is the following:
```
cabal new-exec db-validator -- --db /path/to/db cardano 
--genesisHash 5f20df933584822601f9e3f8c024eb5eb252fe8cefb24d1317dc3d432e940ebb 
--configByron /path/to/configuration/mainnet/mainnet-byron-genesis.json --threshold 0.22 
--configShelley /path/to/configuration/mainnet/mainnet-shelley-genesis.json 
--nonce 1a3be38bcbb7911969283716ad7aa550250226b76a61fc51cc9a9a35d9276d81 --verbose

```

Co-authored-by: kderme <[email protected]>

Add a random cookie to keep-alive messages so that a rouge responder
can't send a MsgKeepAliveResponse before it has read the MsgKeepAlive.

Add a random cookie to keep-alive messages so that a rouge responder
can't send a MsgKeepAliveResponse before it has read the MsgKeepAlive.

Add a random cookie to keep-alive messages so that a rouge responder
can't send a MsgKeepAliveResponse before it has read the MsgKeepAlive.

Give peers with outstanding bytes an advantage when ordering peers.

Fixes #2480.

Return the block size instead of the body size for Shelley.

Take CBOR-in-CBOR and the EBB tag into account for Byron.

Include a test in the serialisation roundtrip test skeleton.

Give peers with outstanding bytes an advantage when ordering peers.

2445: Update index-state and nix updates r=nc6 a=newhoggy



Co-authored-by: John Ky <[email protected]>

Specifically, PR input-output-hk/cardano-ledger-specs#1731 changed the Shelley
transaction generator to no longer fail with QC.discard and moreover to attempt
a given number of retries before failing fatally. That supplants our temporary
workaround.

This commit integrates with the upstream PRs
input-output-hk/cardano-ledger-specs#1735 and
input-output-hk/cardano-ledger-specs#1739, which generalize the contexts and
instances necessary for using the cardano-ledger-specs test case generators.

`tipSampleClient` - `TipRegistry` integration test.

Peer selection which can request local & public root peers but cannot do
gossip.

'Ouroboros.Network.PeerSelection.Governor.PeerStateActions' implement
`PeerStateActions` actions using connection manager.

It's good to keep peer state changes callbacks in a seprate recrod, they
will likely not change much as we go towards decentralisation.

On outound connection we need to be able to manage responders: not only
start them but also re-start them when they are finished.  The server
knows how to do it! This patch provides `ControlChannel`, which allows
to notifiy the server about a thread that it needs to monitor (start
& re-start responders when they finish).

This patch also extends the tests, which now runs multiple rounds of
`ReqResp` protocol, which forces the server to use the restarting logic.

* unidirectional: two connection managers; first, running in initiator
  only mode, connects to the other.
* bidirectional: two connection manager using a single bidirectional TCP
  connection

It is useful to get the 'ConnectionId'.  This will allow to have better
trace messages in the governor integration.