fix: migrate Nebius alias to Token Factory

[demianarc]

What changed

• point the existing nebius alias at the Nebius Token Factory inference endpoint
• add a regression test that locks the alias to the Token Factory URL and existing API-key variable
• refresh the provider guide and example with Token Factory branding and a current model ID
• replace the static model table with Docker Agent's live catalog command and the Token Factory console

Why

Nebius AI Studio evolved into Nebius Token Factory. The old api.studio.nebius.com endpoint and the provider page's sample catalog were left behind, so the built-in alias sent new users to a legacy endpoint and the example referenced a retired model.

This repairs the existing alias; it does not add another provider or generated model catalog. The provider remains on Docker Agent's OpenAI-compatible Chat Completions path.

Nebius migration reference: https://docs.tokenfactory.nebius.com/other-capabilities/migration-guide

Impact

Configurations using provider: nebius now target https://api.tokenfactory.nebius.com/v1 while keeping the same NEBIUS_API_KEY contract. Users get current setup guidance and can discover the catalog without another hard-coded model list becoming stale.

Validation

• go test ./pkg/model/provider -run 'Test(LookupAlias|CatalogAliases|NebiusAliasUsesTokenFactoryEndpoint)$' -count=1
• go test ./...
• ./scripts/build.sh
• go run ./lint . (1,793 files, no offenses)
• go mod tidy -diff
• npx --yes markdownlint-cli2@0.22.1 providers/nebius/index.md (118 files, 0 errors)
• docker-agent models --all --provider nebius --format json (includes openai/gpt-oss-120b)
• git diff --check

Live inference was not run because no Token Factory API key was available in the test environment.