Run reproducer workloads under scoped fault-injection settings

## What to build

Add a run-owned workflow that applies scoped fault-injection settings, executes a reproducer, captures output/artifacts, and restores fault-injection state. This gives agents a safe "run this workload under allocation/slab failure" loop without leaving the guest in a surprising state.

## Acceptance criteria

- [ ] A contributor can run a reproducer under one supported fault-injection profile for a booted Run.
- [ ] Fault-injection state is restored on success, non-zero workload exit, timeout, and transport failure.
- [ ] The response records the applied settings, workload result, and artifact refs.
- [ ] Cleanup failures are surfaced clearly and do not hide the workload result.
- [ ] Tests cover success, workload failure, timeout, cleanup failure, unsupported knob, and artifact capture.

## Blocked by

- https://github.com/randomparity/kdive/issues/910
- https://github.com/randomparity/kdive/issues/918

The issue names no files or tests. Start by reviewing blocked issues #910 and #918, then trace how a booted Run launches a reproducer and captures artifacts. Done means supported profiles run with state restored across every listed outcome, with workload results and cleanup failures both represented in the response.

Written by the indexing model from the issue text.