Update provider model registry and thinking support

[guima-why]

Summary

• Refresh provider model registries for DashScope, DashScope Token Plan, OpenAI/Azure OpenAI, Anthropic, Gemini, Kimi, MiniMax, and Zhipu Coding Plan.
• Add thinking-family support and provider-specific thinking request payloads for Kimi, MiniMax, and Zhipu.
• Add regression coverage for the researched model catalog updates and update auth/model metadata expectations.

Notes

• Bailian models that are scheduled for removal are intentionally retained until they are actually removed upstream.
• Kimi now defaults to kimi-k2.7-code.
• Zhipu glm-5.2 is exposed for the Coding Plan provider, while the non-coding plan providers remain on the currently supported catalog.
• The local research document under docs/ is not included in this commit because the repository ignores /docs/.

Validation

• .venv/bin/python -m pytest tests/providers/test_provider_model_research_updates.py tests/providers/test_thinking_registry.py tests/providers/test_new_providers.py tests/providers/test_openai_provider.py tests/providers/test_dashscope_provider.py tests/providers/test_anthropic_provider.py tests/commands/test_auth_basics.py tests/commands/test_model.py tests/commands/test_effort.py tests/ui/dialogs/test_model_picker.py -q (208 passed)
• .venv/bin/python -m ruff check src tests
• PATH="$HOME/.local/bin:$PATH" git commit ... ran the repository lint and format hooks successfully.