Document parallel short-circuit / early termination pattern

### Problem

A common use case is running N parallel validation tasks where any single "reject" result should immediately short-circuit the remaining tasks. This is different from `Promise.race()` (first to finish wins) — it is about **early termination based on a business logic outcome**.

`CompletionConfig` and completion strategies can be used here, and the `toleratedFailureCount` parameter can serve as a workaround (treating a reject as a thrown error to trigger early completion), but this pattern is not documented.

### Current state

The [parallel operations docs](https://docs.aws.amazon.com/durable-execution/sdk-reference/operations/parallel/#completion-strategies) document `CompletionConfig` and mechanical completion modes, but there is no recipe or pattern for "short-circuit on business logic." The `toleratedFailureCount` workaround is not documented as a pattern.

### Suggested improvement

- Add a pattern/recipe to the best practices or patterns docs: **"Early termination of parallel tasks based on business logic"**
- Document the `toleratedFailureCount` workaround with a concrete example (e.g., a content validation engine with checks that may involve long-running IO)
- Consider linking to the `parallel-race` CDK example if publicly available