feat(i18n): strengthen output-language.md template to enforce language compliance

[zhangxy-zju]

Summary

• Strengthen the output-language.md template wording from soft preference ("Prefer responding") to mandatory rule ("You MUST always respond"), resolving inconsistent output language in multi-turn conversations
• This fixes two classes of language inconsistency:
  • LLM mirroring user's input language instead of the configured output language
  • LLM being influenced by mixed-language content from QWEN.md, downloaded Skills, Agents, and Extensions, causing mid-session language switching or Chinese-English mixed responses
• Add an exception clause allowing users to explicitly override the language for the remainder of a conversation
• Remove the "When a conflict exists" escape clause that allowed injected context to silently override the language setting
• Add 3 unit tests validating the new template content

Changes

File	Change
packages/cli/src/utils/languageUtils.ts	Update generateOutputLanguageFileContent() template
packages/cli/src/utils/languageUtils.test.ts	Add 3 tests for new template wording

Before

## Goal
Prefer responding in **${language}** for normal assistant messages and explanations.

## When a conflict exists
If higher-priority instructions (system/developer) require a different behavior, follow them.

After

## Rule
You MUST always respond in **${language}** regardless of the user's input language.
This is a mandatory requirement, not a preference.

## Exception
If the user **explicitly** requests a response in a specific language, switch to the
user's requested language for the remainder of the conversation.

Test Plan

• npx vitest run src/utils/languageUtils.test.ts — 41 tests pass (3 new)
• Manual: set output-language.md to English, send Chinese input -> LLM should respond in English
• Manual: install an English-language Skill/Agent, set output language to Chinese -> LLM should consistently respond in Chinese without mixing English
• Manual: after language enforcement, say "please reply in English" -> LLM should switch to English for the rest of the session

Linked Issues

Closes #2003

[tanzhenxin]

LGTM!