fix(llm): make request timeout configurable

[guix4ever]

Summary

• Add llm.timeout_seconds with a backward-compatible 60-second default and positive-value validation.
• Expose the setting through everos.toml and EVEROS_LLM__TIMEOUT_SECONDS.
• Pass the configured timeout to both the EverAlgo LLM client and the local OpenAI-compatible provider.
• Update configuration examples, reference documentation, and regression tests.

Fixes #333.

Why

EverAlgo already accepts an LLM request timeout and defaults it to 60 seconds, but EverOS did not expose that value in LLMSettings or pass it through either LLM construction path. As a result, deployments using slower OpenAI-compatible models could not tune the timeout without changing code, even though embedding and rerank clients already expose equivalent settings.

This change keeps the existing default behavior while allowing deployments to adjust the per-request timeout for their model and endpoint.

Area

• Architecture method
• Benchmark
• Use case
• Documentation
• Developer experience
• CI, build, or release

Verification

make lint
make docs-check
pytest tests/unit/test_config/test_settings.py tests/unit/test_component/test_llm/test_client.py tests/unit/test_component/test_llm/test_factory.py -q  # 23 passed
make test         # 1502 passed
make integration  # 78 passed, 6 deselected
make package      # sdist/wheel build and wheel import smoke test passed
git diff --check

Checklist

• I kept the change scoped to the relevant area.
• I am opening this from a separate branch, not pushing directly to main.
• I updated docs, examples, or setup notes when behavior changed.
• I added or updated tests when the change affects behavior.
• I did not commit secrets, .env files, dependency folders, or generated output.
• Active relative links in Markdown files resolve.

Notes for Reviewers

The default remains 60 seconds. Tests verify environment-over-TOML precedence, positive-value validation, and propagation through both LLM construction paths. No real provider request is required for this configuration-only change.

By submitting this pull request, I agree that my contribution is licensed under
the Apache License 2.0.