feat(events): WrappingSet.on/off with mapped values

[jeswr]

What this does

Adds change subscriptions at the collection level: WrappingSet<T> gains on(listener) / off(listener), with listeners receiving the kind of mutation and the mapped JavaScript value rather than a raw quad.

• WrappingSetListener<T> — (event: "add" | "delete", value: T) => void.
• on(listener) subscribes to the underlying dataset's "add"/"delete" quad events, filters them by this set's subject + predicate, and projects the affected quad's object through the termAs mapping the set was created with. Mutations made through any route (the dataset directly, sibling wrappers) are observed.
• off(listener) detaches via a WeakMap<listener, Map<subject␀predicate, adapter>> registry, so it works across the fresh instances SetFrom.subjectPredicate returns on every property access (instance.prop.on(l) … instance.prop.off(l) just works). Re-attaching the same listener for the same subject + predicate replaces the previous subscription, so events are never delivered twice.
• Datasets that don't emit change events are rejected with a new typed DatasetEventsError (extends WrapperError).

Event-emitting datasets are detected structurally (a DatasetCore that is also an EventEmitter with "add"/"delete" quad events, e.g. EventfulDatasetCore from @jeswr/eventful-dataset). That package is a devDependency only (tests + a type-only import); the emitted declarations do not reference it and the package keeps zero runtime dependencies.

Relationship to other PRs

• Supersedes the WrappingSet listener slice of Omnibus #73 — WrappingSetListener, WrappingSet.on/off, and the WeakMap adapter registry keyed by listener → subject␀predicate, ported onto main without the rest of the omnibus (NotificationsDatasetCore is replaced by the structural eventful-dataset surface). Omnibus #73's set-event behaviour is covered by the ported tests in test/unit/wrapping_set_events.test.ts.
• Overlaps with feat: export WrappingSet and narrow SetFrom's return type to WrappingSet<T> #92 on exactly two hunks (exporting WrappingSet from mod.ts and narrowing SetFrom.subjectPredicate's return type to WrappingSet<T>); they are textually identical there, so they drop out on rebase once feat: export WrappingSet and narrow SetFrom's return type to WrappingSet<T> #92 lands. This PR is self-contained against main either way.

Tests

14 new tests (test/unit/wrapping_set_events.test.ts): add/delete/clear emission with mapped values (term and literal mappings), subject/predicate filtering, direct dataset mutations, no event on ineffective adds, multiple independent listeners, live re-iteration inside a listener, off across fresh instances / unknown listeners / per-(subject, predicate) detachment, re-subscription replacement, and the DatasetEventsError path. npm test (136 tests, 0 fail), npx typedoc, and npm audit --omit=dev all pass locally.

Review timing: This draft was prepared with Claude; I (@jeswr) will personally review it before it progresses. I'm currently batching a lot of work in flight, so expect active review Wednesday-Friday (8-10 July).