feat: Retryable operation

[SilanHe]

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

Issue Link, if available

N/A

Description

Add a public retryOperation helper that retries any durable operation (waitForCallback, invoke, waitForCondition, etc.) end-to-end with configurable backoff. Today users have to hand-roll a while loop with manual context.wait() backoff calls. This helper provides a single, replay-safe, well-tested primitive for that pattern, reusing the exact same RetryStrategy shape developers already know from StepConfig.

New files:

• RetryableOperation<T> (sdk/.../util/) — @FunctionalInterface that takes (DurableContext, int attempt) and returns T. Enables lambda syntax for the operation to retry.
• RetryOperationConfig (sdk/.../config/) — Builder-pattern config holding the RetryStrategy (required, same type as StepConfig) and wrapInChildContext (defaults to true).
• RetryOperationHelper (sdk/.../util/) — Static utility with two overloads:
  • Named form retryOperation(context, name, operation, config) — wraps in runInChildContext by default for clean execution history grouping.
  • Anonymous form retryOperation(context, operation, config) — runs directly in the caller's context, no wrapping.

Key behaviors:

• Reuses the existing RetryStrategy / RetryDecision types — zero new retry concepts.
• Replay-safe by construction: every side-effect in the loop is a durable operation (operation.execute() calls durable primitives, context.wait() handles backoff).
• Correctly propagates SuspendExecutionException and UnrecoverableDurableExecutionException without retrying — these are internal SDK control flow signals.
• Default 1-second backoff when RetryDecision.delay() is Duration.ZERO.

Demo/Screenshots

N/A — utility library change, no UI.

Checklist

• I have filled out every section of the PR template
• I have thoroughly tested this change

Testing

Unit Tests

Yes. 36 unit tests across 3 test classes in sdk/src/test/:

• RetryOperationConfigTest (8 tests) — builder validation, defaults, required fields, chaining.
• RetryOperationHelperTest (24 tests in 3 nested classes):
  • NamedForm (9) — child-context wrapping, opt-out, backoff wait naming, default delay, null validation.
  • AnonymousForm (7) — no wrapping, anonymous backoff naming, exception identity, null validation.
  • RetryBehavior (8) — attempt numbers, error forwarding, custom delays, context passthrough, SuspendExecutionException propagation, UnrecoverableDurableExecutionException propagation, retry exhaustion.
• RetryableOperationTest (4 tests) — lambda implementation, context/attempt passthrough, exception propagation, null return.

Integration Tests

Yes. 12 integration tests across 2 test classes in sdk-integration-tests/src/test/:

• RetryInvokeIntegrationTest (6 tests) — first-attempt success, retry after failure, all retries exhausted, custom backoff delays, composition with steps, original exception type preservation.
• RetryWaitForCallbackIntegrationTest (6 tests) — first-attempt success, retry after callback failure, all retries exhausted, composition with steps, multiple failures then success, submitter re-execution on each retry.

Examples

Yes. 2 example handlers with 8 example tests in examples/:

• RetryInvokeExample + RetryInvokeExampleTest (4 tests) — demonstrates retryOperation wrapping context.invoke with retry on transient failures.
• RetryWaitForCallbackExample + RetryWaitForCallbackExampleTest (4 tests) — demonstrates retryOperation wrapping context.waitForCallback with retry on callback rejection.