Add Chronicle query scenarios and release unexpected snapshot streams in tests

[woksin]

Arc for TypeScript 0.37.0 adds in-memory Chronicle query scenarios, the counterpart of the query scenario seeding in Arc for .NET, and aligns snapshot queries that return streams with .NET. npm publication remains disabled.

Added

• ChronicleQueryScenario.for(Model, 'method', ...artifacts) runs a model-bound query against Chronicle read models without a kernel. Seed state per event source with given.forEventSource(id, tenant?): .events(...) builds reducer-backed read models with the Chronicle SDK's ReadModelScenario, and .readModel(instance) pins an instance, which wins over seeded events. Reads are tenant-scoped, and an unseeded key resolves as missing.
  • Listing, watching and projection-backed history need ChronicleKernelScenario, and the scenario says so rather than returning a partial answer.
• snapshotStreamSource(result) in @cratis/arc.core returns the stream a snapshot query returned by mistake, so a test host can release it.

Changed

• A query declared as a snapshot that returns an observable, subscribable or async iterable is rejected without touching the returned source, so a shared RxJS Subject stays usable.
• QueryScenario releases such a stream before rejecting, with a one-second limit and cancellation support. Sources returned inside a scenario are treated as owned by the scenario and may be closed.
• QueryScenario.perform() now rejects queries declared observable: true, as QueryScenario does in Arc for .NET; use ObservableQueryScenario for them.
• yarn ci now type-checks the built declarations of every published package with skipLibCheck: false, so a declaration a consumer can't compile fails the build.

Refs #77