fix: improve model router error handling and add debug logging

fix: handle provider/model format in router (strip prefix before calling provider)

test: comprehensive AI provider testing with 90% success rate

feat: update model router with NVIDIA support and latest model scores

fix: add list_providers method and agents list endpoint

fix: replace deprecated datetime.utcnow with timezone-aware

feat: add multi-model LLM router with providers