Add MiniMax provider models and endpoints

[octo-patch]

Reason: add target provider/model to existing provider registry.

Summary:

• Add MiniMax M3 and M2.7 to the existing model registry with target context, modality, thinking, and pricing metadata.
• Route MiniMax requests through the existing OpenAI-compatible and Anthropic-compatible LiteLLM adapters with selectable global and China endpoints.
• Expose the provider configuration, example environment variables, frontend metadata fields, and focused tests.

Checks:

• PYTHONPATH=backend .venv/bin/python -m pytest -q backend/tests/test_minimax_provider.py
• .venv/bin/ruff check backend/tests/test_minimax_provider.py
• .venv/bin/ruff check --select F backend/pyspur/api/key_management.py backend/pyspur/nodes/llm/_model_info.py backend/pyspur/nodes/llm/_providers.py backend/pyspur/nodes/llm/_utils.py
• .venv/bin/python -m py_compile backend/pyspur/api/key_management.py backend/pyspur/nodes/llm/_model_info.py backend/pyspur/nodes/llm/_providers.py backend/pyspur/nodes/llm/_utils.py backend/tests/test_minimax_provider.py

[octo-patch]

Corrected the MiniMax-M3 default pricing_usd_per_million_tokens to input 0.6 / output 2.4 / cache_read 0.12 to match the provider's published rates, and updated the test assertions accordingly. Ran the MiniMax provider tests (3 passed).