fix: self-heal pkuseg model files before MFA tokenization

feat: auto-install Japanese tokenizer support in MFA conda environment

fix: drop probability-only dictionary rows before MFA parse

fix: always rewrite MFA dictionaries as tab-delimited

fix: normalize dictionary lines to avoid MFA parse IndexError

fix: clean and validate MFA dictionaries by whitespace tokens

fix: detect callable conda-type tools before MFA install

feat: auto-bootstrap micromamba for cloud MFA setup

fix: prefer conda-run MFA entrypoints and detect missing _kalpy

feat: switch Linux MFA install to official conda-forge method

fix: avoid MFA probe import crash and install kalpy with MFA

fix: resolve MFA module entrypoint and release leaked cloud task slots

fix: align MFA verification with runtime command resolution

第一次小型重构测试

sync: align master with local snapshot without force

fix: FFmpeg 中文编码问题

fix: 本地 GUI 优化

fix: MFA 超时时间从 30 秒增加到 120 秒

fix: MFA 数据目录隔离

fix: 并发上线；音频格式

水平線の遥か上を飛んで征く、

fix: 魔搭你要再不行我要去Google Colab了

fix: pkuseg 模型预下载

fix: MFA 哈希校验强制执行

fix: MFA 字典空行校验，哈希校验

fix: 使用 micromamba 安装 MFA

fix: 确保 MFA 正确安装并验证

fix: 修复云端环境检测逻辑

feat: 添加魔搭创空间部署配置

完成了单音频或选择文件夹 → Silero VAD 切片 → Whisper 转录 → 生成 .lab → MFA 强制对齐 → 简单导出，已可用

MFA集成