Snapshots deserialization failures

[kderme]

On testnet, db-sync fails to deserialise stored Babbage snapshots. This means that upon a restart, it rollbacks to the Alonzo/Babbage boundary. This is not a db-sync issues, since the same issue exists on the node.

db-sync-node:Info:66] [2022-07-05 23:09:57.13 UTC] Blocks deleted
[db-sync-node:Warning:66] [2022-07-05 23:09:58.23 UTC] Failed to parse ledger state file ledger-state/testnet-13-snapshot/62688640-414262795a.lstate with error 'DecoderErrorDeserialiseFailure "Ledger state file" (DeserialiseFailure 31749160 "Decoding TxIx: too many bytes.")'. Deleting it.
[db-sync-node:Warning:66] [2022-07-05 23:09:58.28 UTC] Rolling back further to slot 62510369
[db-sync-node:Info:66] [2022-07-05 23:09:58.30 UTC] Rolling back to slot 62510369, hash d931221f9bc4cae34de422d9f4281a2b0344e86aac6b31eb54e2ee90f44a09b9
[db-sync-node:Info:66] [2022-07-05 23:09:58.31 UTC] Deleting 5405 blocks up to slot 62688640
[db-sync-node:Info:66] [2022-07-05 23:10:01.92 UTC] Blocks deleted

[kderme]

Also in node IntersectMBO/cardano-node#4142

[erikd]

I really wonder why there wasn't a simple round trip test to catch this.

Maybe we should even have a round trip test for this in our repo.