feat(omni,speech): add --list-voices and fix cosyvoice voice ID

[hopelynd]

Summary

Improve voice discovery for bl omni and bl speech synthesize: omni gets --list-voices with an expanded built-in voice table; speech fixes an outdated voice ID and adds a link to the full CosyVoice voice documentation in --list-voices output.

bl omni

• Add --list-voices: no API key required; prints a table of 17 built-in output voices (VOICE ID / name / description / language) and exits
• Change --voice help from an inline enum to pointing at --list-voices, so help stays readable
• Expand the built-in voice table from 6 string IDs to structured VoiceEntry rows (including dialect voices such as Dylan, Sunny, Kiki, etc.); default voice changed to Tina.

bl speech synthesize

• Fix longanhuan → longanhuan_v3 in the cosyvoice-v3-flash built-in list
• Append the official CosyVoice voice documentation link at the end of --list-voices output; include the same link in the USAGE hint when --voice is missing
• Add VOICE_TTS_PAGE in urls.ts as the single source of truth for that link

Docs & tests

• Sync skills/bailian-cli/reference/omni.md and speech.md
• Add an e2e test for omni --list-voices