Prevent attempting to suspend multiple times in the same run

[Aaron1011]

The durable sql itself already checks that a task is currnetly running when we try to suspend it, which makes it
very difficult to write tests for. However, checking in the durable sql is insufficient, since we might have an execution that looks like:

• Tokio task A - calls await_event("foo"), and doesn't propagate the ControlFlow::Suspend error.
• Tokio task B (possibly in another process) - calls emit_event("foo")
• Tokio task C - picks up the now-ready task that was previously suspended
• Tokio task A - calls await_event("bar"), which succeeds, since the task is now running.

By adding a check rust-side, we can be sure that we catch incorrect usage of durable

---

Note

Medium Risk
Introduces new validation that turns repeated suspend attempts into TaskError::Validation and changes the ControlFlow::Suspend variant shape, which may break pattern matches and alter runtime behavior for tasks that previously (incorrectly) suppressed suspend errors.

Overview
Prevents tasks from suspending more than once per execution by tracking a has_suspended flag on TaskContext and rejecting subsequent suspend attempts with a validation error.

Changes ControlFlow::Suspend to carry an internal SuspendMarker, forcing suspension to be constructed through TaskContext (via SuspendMarker::new(self)), and updates the worker to match ControlFlow::Suspend(_) accordingly.

^{Written by Cursor Bugbot for commit 5348eee. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}