Add temperature parameter to LLM.chat()

[namantiwari2002]

Add temperature as an explicit parameter to LLM.chat(), following the same pattern used for reasoning_effort. The parameter is threaded through sync_chat, async_chat, and _model_call before being forwarded to the underlying SDK's model.chat() call.

Valid values are floats in the range [0.0, 2.0]. Passing a value outside this range raises a ValueError immediately, before any network call is made. Passing None (the default) preserves existing behavior.

Fixes #75

How tested

• Two new unit tests added to tests/test_llm.py:
  • test_llm_chat_temperature — confirms the value is forwarded to the SDK call.
  • test_llm_chat_temperature_validation — confirms out-of-range values raise ValueError.
• Existing call-site assertions in test_llm_chat, test_model_override, test_fallback_models, and test_empty_response_retries were updated to include temperature=None.
• Full test suite: PYTHONPATH=src python3 -m pytest tests/test_llm.py -q → 27 passed.