Crash during deserialisation of of protocol magic, when it's too large to fit `int32`.

Whenever nodes are configured with a value of protocol magic too large to fit `int32`, this makes the receiving side crash.

https://github.com/input-output-hk/ouroboros-network/blob/master/ouroboros-consensus/src/Ouroboros/Consensus/Ledger/Byron.hs#L780

```
Sep 12 15:24:40 a-node 3xihbn4f05nyd2q38sk8ylr0xiszww26-unit-script-cardano-node_-start[21794]: cardano-node: ExceptionInLinkedThread "ThreadId 43" annotateBoundary: serialization roundtrip failure: DeserialiseFailure 2 "expected int32"
Sep 12 15:24:40 a-node 3xihbn4f05nyd2q38sk8ylr0xiszww26-unit-script-cardano-node_-start[21794]: CallStack (from HasCallStack):
Sep 12 15:24:40 a-node 3xihbn4f05nyd2q38sk8ylr0xiszww26-unit-script-cardano-node_-start[21794]: error, called at src/Ouroboros/Consensus/Ledger/Byron.hs:780:7 in ouroboros-consensus-0.1.0.0-EEi65UrUq7lCWnkSf1LDCN:Ouroboros.Consensus.Ledger.Byron
```

This is as per:
- `cardano-ledger` 22363162ee08039ec6158c24a0882c9357e9e35e
- `ouroboros-network` 05a62acdd64405e284d54a94d55cda832855a84e

Start in Ouroboros/Consensus/Ledger/Byron.hs around line 780 and inspect the deserialization and serialization roundtrip involved in the reported failure. Use the provided oversized protocol-magic scenario and the referenced cardano-ledger and ouroboros-network revisions; done means the receiving side no longer crashes when the value does not fit int32.

Written by the indexing model from the issue text.