[codex] Add configurable LLM settings

[HUAN2022A]

Summary

• Fixes Inconsistent configuration files and useless LLM settings in FULLGUIDE #142 by adding an llm section to config/base_config.yaml and ConfigManager.
• Uses YAML llm settings as runtime defaults while keeping .env variables as overrides for secrets and local changes.
• Makes temperature and optional max_tokens configurable through YAML or LLM_TEMPERATURE / LLM_MAX_TOKENS.
• Updates .env.example and FULLGUIDE.md to document the precedence.
• Adds focused unittest coverage for config parsing, overrides, and OpenAI request parameters.

Why

FULLGUIDE documented llm overrides, but the code only read LLM_MODEL, LLM_BASE_URL, and LLM_API_KEY from .env. Temperature was hard-coded to 0.3, so runtime config could not affect LLM calls.

Validation

• .venv\Scripts\python.exe -m unittest tests.test_llm_config
• .venv\Scripts\python.exe -m py_compile config\config_loader.py config\init.py utils\call_llm_api.py tests\test_llm_config.py
• Manual smoke test with local .env: LLMCompletionCall returned OK