v0.9.0-jv.4: everything since jv.1 — Turbo/Nano cloning, Chatterbox h…

…e/ru/zh/ja, Kokoro Japanese, blends, inline IPA, model_management, Qwen3 memory fixes, libmecab on every platform

v0.9.0-jv.3 — JustVoice's build 3 on audio.cpp v0.9.0

Everything in v0.9.0-jv.2, and:
Chatterbox Turbo and Nano clone a voice from a package converted from Resemble's own checkpoint
(tools/community_models/chatterbox_turbo/convert_chatterbox_turbo.py); Nano's head count comes
from the GGUF (3865d24).

v0.9.0-jv.2 — JustVoice's build 2 on audio.cpp v0.9.0

Everything in v0.9.0-jv.1, and:
Kokoro: a voice_pack request option, a caller's blend as the style pack (86767da).
Kokoro: inline pronunciations "[word](/phonemes/)" in the text (42db68d).

v0.9.0-jv.1 — JustVoice's build 1 on audio.cpp v0.9.0

A VoxCPM2 clone uses the reference clip's transcript (0acac2b).
The Qwen3 forced aligner and Qwen3-ASR word timestamps are in seconds at the input audio's rate (cf08c14).
Modified-file notices (adedc09).

Merge pull request 0xShug0#759 from 0xShug0/feature/crisperwhisper-2-…

…large

Add CrisperWhisper 2.0 with long-form transcription and alignment

Release v0.9.0

Audio8 TTS performance hotfix

Release v0.8.2

Release v0.8.1

Release v0.8.0