Add an LLM provider to specified project

Adds an LLM provider to specified project. If a provider with the same name exists, it is updated (upsert). The health fields (healthStatus, lastHealthCheckAt, lastLatencyMs) are read-only: the health check job writes them and an update keeps the stored values. NOTE on 'endpoint': provide the BASE address only (e.g. https://api.openai.com/v1). The request path (/chat/completions, /embeddings, /audio/transcriptions, /audio/speech, /images/generations, /responses) is appended automatically per request type — do not put it in 'endpoint'. To override a path, set the connection's own defaultChatCompletionPath / defaultEmbeddingPath / ... fields. A query string embedded in 'endpoint' is relocated after the resolved path, so it no longer breaks the URL. NOTE on 'apiVersion': for AZURE_OPENAI connections it is appended to the real request URL as an 'api-version' query parameter (skipped when one is already present). For ANTHROPIC it is sent as the 'anthropic-version' header instead and never reaches the URL. AWS Bedrock and Google Vertex AI build their address entirely in provider-specific code, so neither note applies to them.