feat: add audio file save support to SpeechSDK and MusicSDK

[NianJiuZst]

Summary

Closes #146

Adds save() methods to SpeechSDK and MusicSDK that decode hex-encoded audio from API responses and write it to disk.

Motivation

Both synthesize() and generate() return hex strings in response.data.audio. Saving these to files currently requires users to:

1. Validate hex format manually
2. Decode with Buffer.from(hex, 'hex')
3. Create output directories
4. Generate filenames
5. Handle disk-full errors

This is boilerplate the CLI already handles. The SDK should too.

Usage

import { MiniMaxSDK } from 'mmx-cli/sdk';

const sdk = new MiniMaxSDK({ apiKey: 'sk-...' });

// Speech
const tts = await sdk.speech.synthesize({ text: 'Hello world!' });
const speechPath = sdk.speech.save(tts);
// → speech_2026-05-13-15-30-00.mp3

// Or with explicit path
sdk.speech.save(tts, '/tmp/greeting.wav', 'wav');

// Music
const music = await sdk.music.generate({ prompt: 'Upbeat pop', lyrics: '...' });
const musicPath = sdk.music.save(music);
// → music_2026-05-13-15-31-00.mp3

Files changed

File	Change
src/sdk/speech/index.ts	+50 lines — save() method + shared hex helpers
src/sdk/music/index.ts	+55 lines — save() method + shared hex helpers
test/sdk/speech.test.ts	+35 lines — 4 save test cases
test/sdk/music.test.ts	+35 lines — 4 save test cases

Screenshots