GH-4511: throw SagaConcurrencyException from the Sqlite lightweight saga storage

[jeremydmiller]

Closes #4511.

The Sqlite lightweight saga storage threw a bare System.Exception when an optimistic-concurrency update touched zero rows:

// before
throw new Exception($"Saga version mismatch for {typeof(T).FullName} with id {id}. Possible concurrent update detected.");

Every other lightweight provider throws SagaConcurrencyException at that point — Postgres, SQL Server, MySQL, Oracle — as do the EF Core, CosmosDB, S3, Azure Blob and Redis saga paths. SagaConcurrencyException derives from JasperFx.ConcurrencyException (GH-3444) so that one OnException<ConcurrencyException>().RetryWithCooldown(...) policy covers saga concurrency failures on every provider.

Wolverine's failure-rule matching is ex is T, so on Sqlite the bare Exception fell straight through that policy to the default MoveToErrorQueue continuation. Same saga, same handler code, a different outcome depending on which persistence package is referenced. Fisher-backed hosts inherit it, since Wolverine.Fisher composes over the Sqlite persistence.

Change

src/Persistence/Wolverine.Sqlite/Sagas/DatabaseSagaSchema.cs now throws SagaConcurrencyException with the same message shape as the other providers, so the wording is uniform too.

Why this was invisible

SqliteTests/Sagas/saga_storage_operations.cs already had a concurrency_exception_when_version_does_not_match test — asserting Should.ThrowAsync<Exception>(), which matches anything that throws at all. It passed over the defect for as long as it has existed. It now asserts the exception type, its ConcurrencyException ancestry (the property the retry policy actually depends on), and the message shape.

Verified as a negative control: with the production change stashed, that test fails (1 of 6); with it applied, all 6 pass.

🤖 Generated with Claude Code

https://claude.ai/code/session_01VDUrBeB4tTnKj4AExCS1nj