Prepare 0.2.0 and fix Compose GPU syntax

[jatinkrmalik]

Summary

• Bump the package and About version to 0.2.0
• Point the public site CTAs at the v0.2.0 tag
• Fix compose.yaml so gpus is a device list Docker Compose v2.32 accepts (Docker Desktop 4.38 on Windows rejected gpus: all)

Does not include #50 (resident models / whisper.cpp worker). That stays its own PR.

After merge

1. Tag v0.2.0 on the merge commit and publish the GitHub Release
2. Site CTAs will 404 until that tag exists, so tag right after merge
3. Smoke on Trooper: docker compose up --detach --build and /health/live

Release notes draft (for the GitHub Release)

Since 0.1.0:

• OpenAI-compatible POST /v1/audio/transcriptions for VocaLinux remote hosts
• WebUI Jinja templates
• WPS style remediation and CI gate on cleaned modules
• Catalog expansion, Roman Hinglish Whisper, CPU Whisper latency / WER prune
• Reliable model integrity downloads
• Unified Docker images with runtime CPU kernel dispatch
• Operator docs and .env.example cleanup
• Transcription job lifecycle consolidation

Privacy and security

• No secrets, recordings, transcripts, session IDs, or private hostnames added
• Existing authentication, upload limits, and storage guarantees preserved

[netlify]

✅ Deploy Preview for voca-gateway canceled.

Name	Link
🔨 Latest commit	64be061
🔍 Latest deploy log	https://app.netlify.com/projects/voca-gateway/deploys/6a9a2f6df4b52d0008ad19a6