jsconsumer: correct the docs left describing the removed Nak disposition

The retry path stopped disposing of messages in fc30c0f, but the exported documentation kept describing what it used to do: the Retry overview said a failed delivery is plain-Nak'd, the wiring example said Retry plain-Naks into the ladder, Settle's contract listed "NakWithDelay for redelivery", the capture-failure and stranded paths said the message is Nak'd and that the broker drops that Nak, DeadLetter said a capture failure Naks, Config.BackOff said Retry plain-Naks into it, and OutcomeRetried said the delivery was Nak'd. Every one now describes leaving the delivery untouched for the server's ack timeout, which is what the code does.

The passages explaining WHY a Nak is wrong are kept — that reasoning is the point, and losing it invites the behaviour back.

TestStartClearsStaleServerBackOff had the same problem one level up: it asserted that adopting a Retry wipes the durable's ladder, framed as the desired migration. Clearing is now a hazard rather than a goal, so it is rewritten as TestStartWritesTheConfiguredLadder — Start writes the declared ladder over a stale one (the interim adoption path), and a nil erases what is there (the reason a declaratively-managed ladder cannot simply be omitted).

Refs ENT-1601, ENT-1535.