[Feature]: Add Prompt Self-Optimization Capability to AgentScope

**Is your feature request related to a problem? Please describe.**
Agent performance often degrades not because of model limitations, but because evolving business requirements require frequent prompt updates across agents, skills, and tools, and manual prompt tuning is inherently imprecise, inefficient, and difficult to evaluate systematically.

**Describe the solution you'd like**
Add a built-in prompt self-optimization mechanism to AgentScope by integrating frameworks like dspy-GEPA. Agent execution logs and a small evaluation dataset can be used to automatically improve and version prompts without model fine-tuning.