feat(providers): enable dynamic model selection and flexible API versioning across all providers

[Nithwin]

Is your feature request related to a problem? Please describe.

Goal

Enable developers to use any AI model (--model <name>) and custom API endpoints/versions across all providers (gemini, ollama, groq, openai, anthropic), removing restrictions on default models or hardcoded /v1 endpoint paths.

Problem

Currently, providers default to specific hardcoded model names (e.g., gpt-4o, claude-3-5-sonnet-latest, llama-3.3-70b-versatile) and hardcode the /v1 or /openai/v1 API prefix. This limits users who want to:

1. Use newly released models immediately without waiting for a WindMist update (e.g., o3-mini, claude-3-opus, qwen2.5:32b).
2. Use custom or proprietary enterprise endpoints (such as Azure OpenAI or custom proxy servers that use different path structures instead of standard /v1).

Describe the solution you'd like

Proposed Solution

1. Dynamic Model CLI Flag: Ensure the CLI supports a --model <model_name> override that overrides the default model in ~/.windmist/config.yaml.
2. Flexible Base URL Handling: Allow base_url in config.yaml to accept exact custom API routes without forcing /v1 when a user provides a full custom path.
3. Model Validation: Add helpful error messages or fallback logic when a provider rejects an unsupported model string.

Describe alternatives you've considered

No response

Additional Context

Additional Context

This will make WindMist completely future-proof for all upcoming open-source and proprietary LLM releases across all 5 providers.