Summary

#1657 dropped the preset abstraction and exposed reasoning effort directly, keeping `max` as a DeepSeek-only extension that users opt into knowing standard OpenAI / vLLM / Azure reject it with 400. The TUI still advertised `max` everywhere — `/effort` argsHint, slash-arg picker, ModelPicker effort rows, /effort handler — even when the active endpoint was a third-party host that can't accept it. Users on those endpoints saw `max` in every suggestion and reported it as a preset-era leftover.

Repro (from #1794)

1. Configure baseUrl to a non-DeepSeek endpoint (vLLM, Azure, OpenAI-compat)
2. Run `reasonix code`, type `/` to open the slash dropdown
3. See `/effort <low|medium|high|max>` — `max` would 400 the next call

Fix

New helper `src/cli/ui/effort-choices.ts` exposes `effortChoicesForBaseUrl(baseUrl)` (uses the existing `isDeepSeekHost` allow-list) — returns the full 4-value set on `api.deepseek.com`, drops `max` everywhere else. Wired through:

• Slash autocomplete dropdown — `useCompletionPickers` rewrites the `/effort` spec so argsHint shows `<low|medium|high>` and argCompleter omits `max`
• Slash-arg picker (when user types `/effort `) — same rewrite
• ModelPicker — `effortChoices` prop drives the rendered rows
• /effort handler — rejects `max` with the standard usage hint
• New `effortUsageNoMax` i18n key (EN / zh-CN / de) so the error message on bad input doesn't itself name `max` as an option

`max` stays available on DeepSeek endpoints — that's the design from #1657, just no longer visible on hosts where it would crash.

Test plan

• `tests/effort-choices.test.ts` — 5 cases (DeepSeek allow, third-party filter, null/empty, spoofing rejection, argsHint format)
• `tests/ui-model-picker.test.tsx` — new "hides max on non-DeepSeek" + existing 9 still pass
• `tests/slash.test.ts` — new "rejects max on non-DeepSeek" + "status omits max on non-DeepSeek" + existing 132 still pass
• `npx tsc --noEmit` clean
• `npm run verify` passes (lint + comment policy + full suite)

Fixes #1794