* Record unambiguous failed activity step for future redrive

Only mark the failed sequence when an activity failure propagates. A workflow exception after an activity succeeds must not imply that the activity should be retried.

* test(v2): cover replay of completed activity prefix for redrive

* fix(v2): mark activity redrive boundary only for propagated failure

* feat(v2): validate failed-run activity replay source

* test(v2): qualify redrive plan against persisted failure

* Draft embedded failed-run redrive with reused activity provenance

* Preserve redrive replay time across query reconstruction

* Preserve redrive lineage through embedded history import

* Prove redriven run survives source history expiry

* Preserve redrive replay clock in cold fiber runs

* Verify service worker activity-failure redrive boundary

* Preserve uncaught activity failure identity in worker commands

* Require recorded worker definition for service redrive

* Mark worker-owned workflow definition identity in history

* Preserve workflow timeouts across failed-run redrive

* Format redrive timeout handling

* Reject redrive when activity history advances past failure

---------

Co-authored-by: Durable Workflow <support@durable-workflow.com>