[codex] Align JCode OpenAI base URL with Codex config

[Cloudymap1e]

Summary

• read the active Codex CLI model_provider from ~/.codex/config.toml
• when that provider uses wire_api = "responses" and requires_openai_auth = true, route JCode OpenAI Responses traffic to its base_url
• apply the same base URL override to sidecar OpenAI calls
• add regression coverage for config parsing and response URL selection
• keep session reasoning_effort persisted so overnight session cloning compiles with the existing field usage
• update OpenRouter test constructors for existing provider fields

Why

JCode could trust Codex auth, but its OpenAI provider still hard-coded OpenAI/ChatGPT base URLs. If Codex was configured to use a Responses-compatible gateway, JCode would send the stale API key to the wrong endpoint and fail with invalid key errors.

Impact

This enables JCode to reuse any active Codex-configured HTTP(S) Responses-compatible base URL that explicitly requires OpenAI auth. It does not treat arbitrary chat-completions providers as OpenAI Responses endpoints.

Validation

• cargo test configured_responses_base_url
• cargo test responses_url_uses_active_codex_provider_base_url

---

^{Need help on this PR? Tag @codesmith with what you need.}

• Let Codesmith autofix CI failures and bot reviews