docs(client): drop the ADR-0028 citation from the models example

ADR-0028 in this repository is Metadata Naming & Namespace Isolation
(Deferred) — it says nothing about a model allowlist or plan tiers, and no
ADR in docs/adr/ covers one. The citation was inherited verbatim from the
docblock chain rather than checked, which is the same failure this card
exists to correct.

"plan-filtered" went with it: `AiModelsResponseSchema` declares `models` as
a union of bare id strings and `{ id, label, default }` objects, and states
that both shapes are live — objects when the service exposes the allowlist,
bare ids when it falls back to the adapter's `listModels()`. Both are pinned
accepted in `protocol.test.ts`. The comment now says only what the schema
asserts unconditionally.

Comment text only; no code, no other line, no changeset change.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01D47qPfEWVPmhguWgBZCi5N

Author: claude

packages/client/README.md

@@ -265,7 +265,7 @@ await client.notifications.markRead(['notif-1', 'notif-2']);
 const answer = await client.ai.chat({ messages: [{ role: 'user', content: 'How many open orders this quarter?' }] });
 console.log(answer.content, answer.usage?.totalTokens);
 await client.ai.complete({ prompt: 'Summarise this account in one line:' });
-await client.ai.models();                          // plan-filtered picker list (ADR-0028)
+await client.ai.models();                          // picker list — allowlist objects or bare ids, both live
 await client.ai.conversations.list({ limit: 20 });
 await client.ai.agents.chat('build', { messages: [{ role: 'user', content: 'Draft a follow-up' }] });
 await client.ai.pendingActions.list({ status: 'pending' });   // human-in-the-loop queue