v1.0.2 — Voice discovery features

- voices pick: Interactive voice selector with bubbletea TUI
- voices random: Random voice from filtered pool
- voices tournament: Head-to-head voice elimination rounds

Fix emotion_preset bug and expand test coverage

- Fix: emotion_type was only set for ssfm-v30 when using --emotion preset;
  now correctly set for all models (ssfm-v21 also requires emotion_type: "preset")
- Add buildTTSRequest tests covering all payload fields: volume, pitch, tempo,
  format, seed, language, prev/next-text, emotion presets, output nil defaults
- Add integration tests for --out file saving and format inference from extension
- Fix resetFlags() setting tempo sentinel to "0" instead of "-1", causing
  spurious validation failures in tests
- Add base-url to config keys and wire NewWithBaseURL in root/voices commands

Co-Authored-By: Claude Sonnet 4.6 <[email protected]>

Initial release of cast — Typecast TTS CLI