Cardano Updates

Signed-off-by: Sasha Bogicevic <[email protected]>

Optimizations applied:
 * Single-pass serialization: each 'TxOut' is serialized once via
   @serialiseData . toBuiltinData@ and reused for both UTxO hash checks
   and accumulator scalar computation.
 * Combined CRS lookup: the reference input for the CRS is looked up once
   and decoded only when the membership check is actually reached.

Signed-off-by: Sasha Bogicevic <[email protected]>

Signed-off-by: Sasha Bogicevic <[email protected]>

This should be used as a reference input to reduce the execution
of contracts that need CRS string.

Signed-off-by: Sasha Bogicevic <[email protected]>

Hook new crs validator to script registry

Signed-off-by: Sasha Bogicevic <[email protected]>

Remove CRS from Fanout redeemer

Signed-off-by: Sasha Bogicevic <[email protected]>

Create BLS accumulator benchmarks with UTxO

Signed-off-by: Sasha Bogicevic <[email protected]>

Signed-off-by: Sasha Bogicevic <[email protected]>

Signed-off-by: Sasha Bogicevic <[email protected]>

Signed-off-by: Sasha Bogicevic <[email protected]>

Hash TxOuts to get a 32 bytes string length

Signed-off-by: Sasha Bogicevic <[email protected]>

Fix the failing observer test

Related log book entry is here https://github.com/cardano-scaling/hydra/wiki/Logbook-2025-H1#2025-12-01

Signed-off-by: Sasha Bogicevic <[email protected]>

Temporary add acc fn to weeder ignore

Remove utxoHash from Snapshot

Remove leftover fn from IsTx class

  Hide internal details (flush, reinitLastSeen, Connection, migration)
  from the public withSQLiteEventStore callback. The callback now receives
  only an EventStore — flush is handled internally by sourceEvents
  (auto-flush before reads), rotate, and the bracket cleanup; migration
  runs automatically on startup given the legacy state file path.

  Add schema versioning using SQLite's PRAGMA user_version. Migrations
  are applied incrementally on startup via migrateStep, making future
  schema changes straightforward. The node refuses to start if the
  database version is newer than expected (downgrade protection).

  Extract all inline SQL into named query functions for readability.

  Add migration tests: fresh DB version check, idempotency, and
  downgrade rejection. Update architecture docs with SQLite schema,
  async write-behind, and schema versioning documentation.

  Adds withSQLiteEventStore that manages the writer thread lifecycle
  automatically — link surfaces crashes, finally ensures flush + cancel
  on exit. Callers no longer need to manage cleanup manually.

  Move SQLite event writes off the hot path using a bounded TBQueue
  (capacity 1000) and a single background writer thread. Events are
  JSON-encoded eagerly and enqueued; the writer drains the queue and
  batch-inserts via executeMany in a single transaction.

  Key changes:
  - EventSink gains a putEvents field for batch writes; mkEventSink
    provides a default sequential fallback
  - putEventsToSinks switches from event-major to sink-major loop order
  - SQLiteBased uses BLOB storage instead of TEXT, WAL mode with
    synchronous=NORMAL, and 64 MB page cache
  - Flush marker pattern (TMVar) lets rotate and tests block until all
    queued writes hit disk
  - Migration bypasses the async queue and writes directly via executeMany
  - Node replay collects events into a list then calls putEventsToSinks
    once instead of per-event mapM_
  - Rotation module delegates to underlying putEvents for batch support

Got all changes from the capi-10.21 branch here