Add OpenRouter Support

[crmne]

Context

I'm hitting Anthropic rate limits constantly with my own usage, and several community members have requested OpenRouter integration. This would allow access to a wider range of models through a single API key while maintaining RubyLLM's unified interface.

Benefits

• Single API key to access models across providers
• Potential cost savings through OpenRouter's pricing
• Simplified rate limit management
• Access to exclusive models not available through direct provider integrations
• Fallback capabilities when primary providers are at capacity

Implementation Considerations

• OpenRouter largely follows OpenAI's API structure, so we can likely adapt our existing OpenAI provider implementation
• Need to handle model ID mapping/translation to keep the model selection experience consistent
• Should implement proper error handling for OpenRouter-specific cases
• Will need to update the Models registry to include OpenRouter-accessible models

Scope

Initial implementation should focus on:

• Chat completion support (highest priority)
• Embeddings support
• Model listing
• Full streaming support
• Tool use

Image generation through DALL-E can be a second phase.

Next Steps

• Research OpenRouter API documentation
• Create a new Provider module for OpenRouter
• Implement proper model discovery and capabilities
• Test with complex conversations, tools, and streaming