Implement Nebius chat completions adapter

[Photon101]

Summary

• replace the Nebius AI stub with the documented Token Factory /v1/chat/completions integration
• use the current Token Factory API base and preserve the existing Llama 3.3 default model
• send system/user messages, max token and temperature options, plus provider-specific extra fields
• normalize assistant-message and text-style choices plus prompt/completion token usage
• expand tests from smoke-only to dry-run, request-shape, text-choice fallback, usage mapping, and error coverage

Validation

• corepack pnpm exec vitest run packages/ai/nebius/src/index.test.ts
• corepack pnpm exec tsc -p packages/ai/nebius/tsconfig.json --noEmit
• git diff --check

Reference

• Nebius chat completions docs: https://docs.tokenfactory.nebius.com/api-reference/inference/create-chat-completion
• Nebius quickstart: https://docs.tokenfactory.nebius.com/quickstart