Move cluster-induced workflow shutdown from engine into web

### Task Summary

`amber/src/main/scala/org/apache/texera/amber/clustering/ClusterListener.scala` listens to Pekko cluster events and, when a node leaves, iterates `WorkflowService.getAllWorkflowServices` to call its own `forcefullyStop(executionService, cause)` on each in-flight execution. Both calls reach into web (`web.service.WorkflowService`, `web.service.WorkflowExecutionService`, `web.storage.ExecutionStateStore`).

After the engine event work lands, `ClusterListener` only needs to publish an engine event (e.g. `ClusterNodeRemoved`); the web-side subscriber handles enumeration of active workflows and the force-stop business logic. Move the body of `forcefullyStop` and the `WorkflowService` iteration to that web-side subscriber. `ClusterListener` ends up as a thin "Pekko cluster → engine event" bridge with no web imports.

Part of #5424. Depends on the engine-event proto sub-issue.

### Task Type

- [x] Refactor / Cleanup