fix(scheduler): correct claim retry queue (ladder, give-up, native)

[ephpm-claude]

The claim retry queue merged in #104 ships three correctness bugs. This patches them in place on current main (the queue is config-gated but enabled by default, so it runs on every host). Supersedes my earlier #103, which went stale once #104/#105 moved main.

1. Ladder never advances / give-up never fires. fireDue deleted the popped item from the index, so runOne's failure re-Add built a fresh item — attempts reset to 1 (stuck at ~30s) and firstFailure reset every fire (MaxAge never triggered). An unclaimable job retried every 30s forever, burning the very shared token it was waiting on. Fix: fireDue keeps the item in the index across a fire.
2. Retries dropped after one attempt. retryHandler always returned nil, and runOne dropping on nil removed the item handleQueued had just re-Added synchronously. Fix: retryHandler returns the real claim error via a context pointer that enqueueRetryIfEligible fills in (also suppressing the duplicate enqueue); the rate-limit class is preserved so snap-to-reset still works.
3. Native macОS didn't self-heal — handleNativeMacOSJob was the one claim path not wired to the queue. Wired it up.

Tests

• TestRetryQueue_StillFailingRetryStaysEnqueued drives the real fireDue→retryHandler→claim path and fails before this change.
• TestRetryQueue_LadderAdvancesThroughFirePath asserts the delay grows 30s→1m→2m and firstFailure is preserved.
• Full scheduler suite green, -race and go vet clean.

🤖 Generated with Claude Code