Add OpenAI-compatible inference relay with model-bound profiles, tokens, and providers

Hive can now act as an OpenAI-compatible inference relay. You can create model-bound profiles and tokens, manage upstream providers from the dashboard, and route relay requests through a runtime-managed or environment-backed provider. Each relay request records input tokens, output tokens, estimated cost, and audit activity. The operations dashboard adds Inference > Profiles and Providers, profile and token detail pages, usage widgets, period-aware charts, and profile-level client configuration. Seed data creates sample providers, profiles, tokens, and usage so the dashboard has realistic data locally. This feature is opt-in and does not change existing agent settings.