Custom queue-storage schema: reject runtime slot counts that differ from the installed ring #689

A runtime on a custom queue-storage schema can be configured with different ring slot counts from the ones the schema was installed with, and nothing rejects it.

How it happens. install_queue_storage_substrate seeds {queue,lease,claim}_ring_state.slot_count with ON CONFLICT (singleton) DO NOTHING, so preparing an existing schema with different counts keeps the installed counts without error. Rotation then follows the installed slot_count, while several runtime paths bound slots by the configured count:

• complete_receipt_runtime_batch_fast maps the claim slot through ring_slot_index(.., claim_slot_count(), ..), so a claim in a slot at or above the configured count fails completion with invalid claim ring slot N.
• Both receipt-rescue loops iterate 0..claim_slot_count(), so claims in higher slots are never rescued.
• oldest_initialized_claim_slot_tx filters slots to < claim_slot_count().

With fewer configured than installed slots, jobs claimed into the higher slots can fail to complete or stay stranded after a worker dies. The default awa schema is not affected: both the installer and QueueStorage::new force default slot counts there.

Proposed fix. At startup (or in prepare_schema), read the three *_ring_state.slot_count values and refuse to run when the configured counts differ, naming both values. Admin reads already use installed slot numbers (#665, #688).