Add OPENAI_LLM_TIMEOUT env var for custom OpenAI-compatible provider

[joseluisll]

Summary

Adds support for OPENAI_LLM_TIMEOUT environment variable to control LLM timeout when using custom OpenAI-compatible providers.

This PR addresses issue #1010.

Changes

run_agent.py

• Modified _build_api_kwargs() method in AIAgent class (line 2624)
• Timeout now dynamically uses OPENAI_LLM_TIMEOUT env var when provider is "custom"
• Falls back to default 900 seconds for other providers or invalid env values

tests/test_run_agent.py

Added comprehensive test coverage:

• test_timeout_custom_provider_with_env: Verifies custom timeout from env var
• test_timeout_custom_provider_without_env: Verifies default timeout when env not set
• test_timeout_openrouter_ignores_env: Confirms other providers ignore the env var
• test_timeout_invalid_env_value: Validates fallback to default on invalid values

hermes_cli/config.py

Added OPENAI_LLM_TIMEOUT to OPTIONAL_ENV_VARS (line 576)

• Marked as advanced setting
• Documents purpose and usage

Behavior

Provider	OPENAI_LLM_TIMEOUT Set	Timeout Used
custom	Yes	Env value
custom	No/Invalid	900 seconds
Other	Any	900 seconds

Testing

All tests passing:

• ✅ test_basic_kwargs
• ✅ test_timeout_custom_provider_with_env
• ✅ test_timeout_custom_provider_without_env
• ✅ test_timeout_openrouter_ignores_env
• ✅ test_timeout_invalid_env_value

[joseluisll]

@teknium1 Can you please review this PR?. A 900s hardcoded delay is a stopper for developers using local LLM´s.

[arilotter]

this commit includes many unrelated formatting changes. please revert those.

[joseluisll]

this commit includes many unrelated formatting changes. please revert those.

I have restored the formatting changes, so only the TIMEOUT remains, for the custom provider.

[joseluisll]

change done by PR: #1194 . Use the environment variable HERMES_API_TIMEOUT