Update desktop inference provider defaults

[pmbstyle]

Summary

• add Minimax to the Desktop inference provider picker
• refresh default model IDs across the shared provider catalog and Desktop setup
• populate provider base URLs when selecting Octo or worker inference providers
• update provider/default model tests

Validation

• npm run build from desktop/
• uv run pytest tests/test_configure_llm.py tests/test_litellm_profile_resolver.py tests/test_openrouter_provider.py tests/test_litellm_provider_payload_hardening.py tests/test_runtime_mcp_integration.py tests/test_settings_config_sync.py
• git diff --check