[Epic] Evolution to AI-Native Observability: Integration of Hera Agent

### Background
Standard observability platforms require manual dashboard checks and expert knowledge to identify root causes. To reduce the MTTR (Mean Time To Repair), we propose evolving Ozhera into an AI-Native platform.

### Goal
Implement a conversational interface (Hera Agent) that can:
- Automatically summarize system health status.
- Perform root cause analysis (RCA) by querying metrics, logs, and traces.
- Execute maintenance tasks via predefined API interfaces (e.g., scale-up/down, log-level adjustment).

### Key Components
- Integration with LLM frameworks (LangGraph/CrewAI).
- Asynchronous execution engine (Virtual Threads support).
- Natural language to PromQL/SQL translation.