fix: don't publish flashblock after payload cancellation

[avalonche]

📝 Summary

Prevents a cancelled payload job from publishing locally built payloads into the payload handler after the job has already been resolved (by newPayload).

It also renames PayloadJobCancellation::cancelled() to wait_for_cancellation() so async wait sites are easier to distinguish from immediate is_cancelled() checks.

💡 Motivation and Context

1. forkchoiceUpdated starts a payload job.
2. The job builds fallback and flashblock payloads.
3. Built payloads are sent through local channels to the PayloadHandler.
4. The handler emits Events::BuiltPayload, which updates reth's engine tree with the local payload.

build_next_flashblock() was a blocking job that previously performed side effects such as publishing. Reth could then see a stale local same-height payload after the job had already been resolved. Now:

• build_next_flashblock() builds and returns BuiltFlashblockOutput.
• The async outer loop owns publication.
• The outer loop checks cancellation after the blocking task returns and before publishing.
• publish_flashblock_payload() performs one final cancellation check before websocket publish and channel sends.

---

✅ I have completed the following steps:

• Run make lint
• Run make test
• Added tests (if applicable)

[julio4]

We were already checking for block cancel before calling all publish side effects in sync, so this is mostly refactor to extract publish side effects outside of build_next_flashblock it seems?

[julio4]

Also please see #518

[akundaz]

Aren't we checking for cancellation just before we call this publish function? So this seems unnecessary here

[avalonche]

I believe we check is_resolved but is_cancelled() can also be true if deadline / error is reached. but yes I believe it can be refactored better