feat: add TTS Model ID configuration UI

[YizukiAme]

Summary

Add a Model ID input field to the TTS provider settings dialog, allowing users to customize the model used for text-to-speech generation (e.g., switching from gpt-4o-mini-tts to tts-1-hd for OpenAI).

Previously, model IDs were hardcoded and users had no way to change them through the UI.

Closes #14

Changes

UI (components/settings/tts-settings.tsx)

• Add Model ID input field with conditional rendering (only shown for providers that support model IDs: OpenAI, GLM, Qwen)
• Import DEFAULT_TTS_MODELS constant for placeholder display and conditional logic
• Default model shown as placeholder text (e.g., gpt-4o-mini-tts for OpenAI)

Constants & Types (lib/audio/)

• Add DEFAULT_TTS_MODELS map in constants.ts with default model IDs per provider
• Add modelId to TTSProviderConfig type definition
• Update TTS providers to read modelId from config with fallback to defaults

API & Integration

• Wire modelId through the TTS API route (app/api/generate/tts/route.ts)
• Pass modelId from scene generator (lib/hooks/use-scene-generator.ts)

i18n (lib/i18n/settings.ts)

• Add ttsModelId translation key for Chinese ("模型 ID") and English ("Model ID")

Test Plan

• Select OpenAI TTS → Model ID field visible with gpt-4o-mini-tts placeholder
• Select GLM TTS → Model ID field visible with emoti-voice placeholder
• Select Azure TTS → Model ID field hidden (Azure doesn't use model IDs)
• Select Browser Native TTS → Model ID field hidden
• Custom model ID persists after page reload
• Empty model ID falls back to default

[YizukiAme]

Closing in favor of #50 which provides a more comprehensive solution covering both TTS and ASR model configuration, with a UI pattern consistent with image-generation model management.