feat: Add glossary support for translation

[huketo]

Summary

Add glossary (용어집) support so that specific terms are consistently translated according to user-defined mappings. Currently, the translator relies entirely on the LLM's judgment for term translation, which can lead to inconsistent or incorrect translations for domain-specific terms, proper nouns, or preferred expressions.

Motivation

• Certain terms (e.g., brand names, technical jargon, community-specific expressions) should always be translated in a specific way
• LLMs may translate the same term differently across posts, leading to inconsistency
• Some terms should be kept untranslated (e.g., proper nouns)

Key Questions to Discuss

1. Storage format: Should the glossary be defined in config.toml, a separate file (e.g., glossary.toml/glossary.json), or both?
2. Scope: Should glossary entries be global or per target language?
   • Global: "블루스카이" → "Bluesky" (same for all languages)
   • Per-language: "블루스카이" → {"en": "Bluesky", "ja": "ブルースカイ"}
3. Integration point: Inject glossary terms into the LLM system prompt (simplest) vs. pre/post-process text replacement (more deterministic)?
4. Size limit: Should we cap the number of glossary entries to avoid bloating the prompt?

Current Translation Flow

The translation prompt is built in internal/translator/prompt.go via buildTranslateSystemPrompt(). The glossary could be appended as an additional instruction section in the system prompt, e.g.:

Glossary (use these exact translations):
- 블루스카이 → Bluesky
- 따라봇 → DDaraBot

Labels

enhancement