fix: improve app name derivation to support numbers and acronyms

[nakrenat]

🚀 Summary

The current PART_FINDER_RE regex in app.py has a limitation: it fails to capture numbers and handles acronyms inconsistently. This PR updates the regex to correctly derive kebab-case app names from PascalCase/CamelCase strings that contain digits or consecutive uppercase letters.

🛠 Changes

• Updated PART_FINDER_RE to: r"[A-Z][a-z]+|[A-Z]+(?=[A-Z][a-z]|$)|[a-z]+|\d+"
• This ensures that digits (like '2' in Sora2Video) and acronyms (like 'ONNX') are properly parsed as distinct parts.

✅ Test Cases (Verified locally)

I've verified the fix with the following cases:

• Sora2Video -> sora-2-video (Previously: sora-video)
• ONNXModel -> onnx-model (Previously: o-n-n-x-model)
• My3DApp -> my-3-d-app (Previously: my-app)

This improvement ensures a better Developer Experience (DX) when naming fal.ai apps.

[nakrenat]

It seems like the Claude Code Review action is failing due to missing id-token: write permissions. This shouldn't affect the core changes in my PR.

[nakrenat]

Hi! It seems like the integration tests are failing with 401 Unauthorized because repository secrets are not shared with this PR from my fork. I've verified the regex fix locally with several test cases (Sora2Video, ONNXModel, My3DApp) and they all produce the correct output now. Could a maintainer trigger the tests or review the logic?