Standardize error messages across all LLM providers for consistency

[himanshu231204]

Description

Different LLM providers (OpenAI, Anthropic, Gemini, Groq, etc.) use different error message formats when things go wrong. This makes debugging harder for users who switch between providers. Error messages should be standardized.

What to Do

1. Review error handling in each LLM provider:
   • \openagent_eval/providers/llm/openai.py\
   • \openagent_eval/providers/llm/anthropic.py\
   • \openagent_eval/providers/llm/gemini.py\
   • \openagent_eval/providers/llm/groq.py\
   • \openagent_eval/providers/llm/openrouter.py\
   • \openagent_eval/providers/llm/ollama.py\
2. Create a consistent error format that includes: provider name, error type, and actionable message
3. Update all providers to use the same format

Files Involved

• \openagent_eval/providers/llm/*.py\ — all LLM providers
• \openagent_eval/exceptions/\ — exception classes

Good First Issue ✅

A code quality task that teaches you the provider architecture.