Add OpenAI gateway API key header support

[Deniz Dalkılıç (dalkilicdeniz)]

Summary

Add opt-in support for OpenAI-compatible gateways that require OPENAI_API_KEY in a provider-specific HTTP header in addition to standard Bearer authentication.

Motivation / linked issue

Some API-management gateways expose an OpenAI-compatible /v1 surface but require credentials in an api-key or subscription-key header. LiteLLM's openai/* provider sends Bearer authentication by default, so ASSERT pipeline, init, and ACS calls currently receive 401 from those gateways.

Changes

• Add ASSERT_OPENAI_API_KEY_HEADER to name the compatibility header while reusing OPENAI_API_KEY as its value.
• Apply the opt-in header consistently to Chat Completions, Responses, assert-ai init, and ACS generation calls.
• Preserve explicit per-call headers and keep existing behavior unchanged when the variable is unset.
• Document the setting in .env.example and add focused regression coverage.

Testing

• python -m pytest -q tests/test_model_client.py tests/test_init_llm.py tests/test_acs_language_model.py (94 passed)
• Live sanitized smoke test: ASSERT structured JSON generation through an APIM /openai/v1 gateway with openai/gpt-54 succeeded.

Checklist

• Tests pass locally (pytest and/or viewer checks as applicable).
• Docs updated if behavior or public API changed.
• No secrets, credentials, or customer data committed.
• No breaking change, or a CHANGELOG.md entry is included.