[carbolymer]

This answers a remark on PR #17, which asked to remove the eons as much as possible and to use ShelleyBasedEra era where appropriate: #17 (comment)

ShelleyBasedEra is now the only era witness in the server. The JSON output does not change.

What changes

• Removes AlonzoBasedEra, IsAlonzoBasedEra and ToShelleyBasedEra. Dijkstra gets its IsShelleyBasedEra instance.
• Wrappers and encoders for values that exist only from Alonzo onwards take ShelleyBasedEra era together with PlutusEraConstraints era. It bundles EraPlutusContext era, EraTxCert era and Script era ~ AlonzoScript era.
• Matches on those values list only the four Alonzo-onwards eras and have no error arms. The script equality makes GHC treat the pre-Alonzo arms as inaccessible, so writing one is a compile error.
• PlutusEraConstraints is a plain constraint synonym, so no continuation-passing provider function is needed.
• Adding an era now means one ShelleyBasedEra constructor, and -Werror flags every match that misses it.
• Fixes four -Wincomplete-patterns warnings in the unit tests. All four are missing Dijkstra cases that already exist on master.
• The test generator for redeemer pointers now covers Dijkstra. This regenerates 12 EvaluateTransactionResponse test vectors, with no schema failures.

Notes for reviewers

• PlutusEraConstraints lives in Ogmios.Prelude, which already disables -Wredundant-constraints.
• PlutusEraConstraints is left off Alonzo's and Babbage's encodeUtxoFailure, where nothing needs it.

Testing

• The nix build passes with -Werror.
• The unit suite passes on the rebased branch: 158 examples, 0 failures.
• After the first two commits, all 737 test vectors came out byte-identical.

There is no user-facing change, so there is no changelog entry.

This PR is stacked on PR #17, so merge it after #17:

• Bump Cardano dependencies and report Dijkstra submission errors #17

[Jimbo4350]

LGTM but consider the change I suggested in the comment.