feat: add MiniMax and OpenAI as configurable LLM providers for generative search

[octo-patch]

Summary

• Refactor the hardcoded Ollama integration in --generative to support configurable LLM providers
• Add MiniMax as a cloud LLM provider via OpenAI-compatible API (api.minimax.io/v1)
• Add OpenAI as an additional cloud provider option
• Support provider auto-detection from MINIMAX_API_KEY / OPENAI_API_KEY environment variables

Changes

File	Change
seagoat/utils/llm_provider.py	New provider abstraction with streaming support for Ollama/OpenAI/MiniMax
seagoat/utils/generative.py	Refactored to use provider module instead of hardcoded Ollama
seagoat/utils/config.py	Extended config schema with generative section (provider/model/apiKey/baseUrl/temperature)
seagoat/cli.py	Pass config to enhance_results()
pyproject.toml	Add openai SDK dependency
README.md	Add provider configuration docs with table

Configuration

# .seagoat.yml
generative:
  provider: minimax        # ollama | openai | minimax
  model: MiniMax-M2.5      # optional

Or just set MINIMAX_API_KEY env var for auto-detection.

Test plan

• 31 unit tests for llm_provider module (provider detection, config, streaming for all 3 providers)
• 13 unit tests for generative module (thinking-tag stripping, result filtering, config passing)
• 9 integration tests (config validation, end-to-end MiniMax flow, provider switching)
• All 53 new tests pass
• All 9 existing config tests pass (no regressions)
• Verify with real Ollama server (backward-compatible, no behavior change for existing users)
• Verify with real MiniMax API key