Add DeepSeek provider support for v0.2.3

[HuRuilizhen]

Summary

This PR is the first 0.2.3 slice for extending the provider-backed LLM extraction path.

The main goal is to add DeepSeek support for the official endpoint and adapt the current output-contract layer so deepseek-chat can participate in the existing LLM extraction flow without changing the public pipeline API.

Included

• adds DeepSeek provider support

  • extends the supported provider set to include deepseek
  • adds the official default base URL https://api.deepseek.com/v1
  • adds the default API key env var DEEPSEEK_API_KEY
  • keeps DeepSeek on the current OpenAI-compatible chat-completions path

• adapts DeepSeek output-contract behavior

  • adds a DeepSeek-specific provider client specialization
  • makes auto prefer json_object before prompt_only for DeepSeek
  • does not change the existing auto-contract sequence for other providers

• adds focused provider coverage

  • verifies DeepSeek default provider wiring
  • verifies DeepSeek auto-contract preference
  • verifies the DeepSeek specialization is selected by the provider builder

• updates the shared LLM example

  • extends examples/llm_extraction.py to show how to run with deepseek-chat
  • keeps one shared example rather than adding a provider-specific duplicate script

Validation

Repository checks:

• pytest -q tests/test_llm_provider.py
• ruff check src/labelgen/config.py src/labelgen/extraction/llm_provider.py tests/test_llm_provider.py examples/llm_extraction.py
• pyright src/labelgen/config.py src/labelgen/extraction/llm_provider.py tests/test_llm_provider.py examples/llm_extraction.py

Notes

• this PR does not yet perform live DeepSeek verification; that will happen after credentials are supplied
• benchmark work remains a separate follow-up branch for 0.2.3
• local planning documents under docs/ remain intentionally untracked and are not part of this branch