fix(speech): correctly enable and download subtitles for TTS

[raylanlin]

Summary

Fixes `--subtitles` flag for `mmx speech synthesize`, which was not working due to two API spec mismatches.

Root Causes

1. Wrong request parameter: Code sent `subtitle: true` but the API expects `subtitle_enable: true`
2. Wrong response field: Code expected inline `subtitle_info` object, but API returns `subtitle_file` — a download URL to a JSON file

Changes

File	Change
`src/types/api.ts`	`subtitle` → `subtitle_enable`; `subtitle_info` → `subtitle_file`; remove dead `SubtitleInfo` type
`src/commands/speech/synthesize.ts`	Fix request param; download subtitle JSON from URL; convert to SRT format; add error handling + quiet mode support
`test/commands/speech/synthesize.test.ts`	Add dry-run test verifying `subtitle_enable: true` in request body
`skill/SKILL.md`	Update `--subtitles` docs with usage example

How It Works

When `--subtitles` is passed:

1. Request sets `subtitle_enable: true`
2. API returns `data.subtitle_file` — an OSS URL to a JSON subtitle file
3. CLI downloads the JSON, converts timestamps to SRT format, and saves alongside the audio file
4. Example: `--out hello.mp3` → saves `hello.mp3` + `hello.srt`

SRT Output Example

```
1
00:00:00,000 --> 00:00:05,438
你好，这是第一条字幕。接着是第二句。

2
00:00:05,580 --> 00:00:06,764
谢谢收听。
```

Testing

• Built locally with `bun run build`
• Tested against live API with multiple Chinese and English texts
• Subtitle file correctly generated with proper SRT timestamps
• Dry-run test verifies correct request parameter name
• All existing tests pass

Closes #102