Cardano Updates

Add ToJSON/FromJSON instances for `NonEmptyMap` and `NonEmptySet`

Added the following tests:

* JSON roundtrip property test
* Unit test for trying to decode an empty object `{}` for the
  `NonEmptyMap`, and an empty list `[]` for the `NonEmptySet`. These
  scenarios are expected to fail with a specific error message.

Using `pre-commit run --all-files` will now run `fourmolu`,
`cabal-gild`, `shellcheck` and `nixpkgs-fmt` on all modified files.

Did not add `hlint` for now because there are *lots* of warnings that
will take time to fix.

Since the UMap purge, ordering matters when it comes to
registering pools, stake credentials and delegations.

`validateOutsideForecast`, `validateOutputTooBigUTxO`

After removing distinct_bytes, the CDDL generator could produce
duplicate empty byte strings for plutus scripts in nonempty_set
collections (tag 258). The Set decoder would deduplicate them,
causing a length mismatch error. Fix by generating 32-byte random
bytestrings which makes collisions virtually impossible.

Also mark proposal_procedure antiCborSpec as pending due to
unrelated protver decoder issue.

Co-Authored-By: Claude Opus 4.6 <[email protected]>

Introduce DijkstraEraBlockHeader to have a lens to the previous nonce,
and the DijkstraBbodySignal type to constrain the header existantially,
same as for BbodySignal.

Move PerasCert, PerasKey and validatePerasCert to dijkstra's BlockBody.

Remove PrevEpochNonceNotPresent because the field is implicitly present
in the signal to the BBODY rule now.

Refactor the dijkstra BBODY rule to reuse validations from shelley,
alonzo, conway, inlined, because the change in the type of Signal makes
calling earlier BBODY transitions impossible.

Export reusable validations to future eras and update the BBODY signal.

Export reusable validations to future eras and update the BBODY signal.

Update version dependencies for core and shelley packages.

Currently, the `B` constructor contains a `ByteString` (pinned in
memory) with a max length of 64 bytes. Since the `ByteString` is small
and can be instantiated frequently, it makes more sense to use an
unpinned byte array representation (`Data.Array.Byte.ByteArray`) so
that we can take advantage of Haskell's garbage collection to improve
memory allocation of objects in the heap.

We also place an upper bound on `nothunks` because of new orphan
instance (`instance NoThunks ByteArray`) that is going to be upstreamed.

We do should *not* provide a default implementation for 'fromPlutusData'
because types should always be invertible from 'PlutusData'. Therefore,
we force the user to make a conscious decision to not provide an
implementation for this function.