daemon: ignore duplicate task exit events in daemon state

[fuweid]

• supersedes / closes Handle duplicate exit events #46391
• ref: libcontainerd: consider to use task.Wait to update the container's exit code instead of task.Event #46212

- What I did

• Added daemon-side duplicate TaskExit filtering based on container state and timestamps
  to avoid reprocessing late exits.

- How I did it

• Introduced shouldIgnoreExitEventWithLock in daemon/monitor.go and invoked
  it from handleContainerExit while holding the container lock.
• Ignored duplicates when the container is removing/exited/dead, or when a
  restart is in progress and the exit timestamp is newer than FinishedAt, or
  when a new task is already running and the exit predates StartedAt.

- How to verify it

• Manually run a workload that triggers duplicate TaskExit events (e.g., shim
  panic during restart) and confirm no state regressions.
  • Manually build containerd and force it to send duplicate TaskExit on task.Delete API call
  • Using failpoint-based shim to run
    • docker run -d --runtime runc-fp --annotation "io.containerd.runtime.v2.shim.failpoint.Delete=1*panic(oops)" ubuntu:latest sleep 5

- Human readable description for the release notes

Fix daemon handling of duplicate container exit events to avoid repeated cleanup and state transitions.

- A picture of a cute animal (not mandatory but encouraged)

REF: #46212

[fuweid]

ping @dmcgowan @AkihiroSuda for review when you have chance. thanks

[lingotesoropuro-jpg]

Gracias

[lingotesoropuro-jpg]

Gracias