Route synthetic data generators through LiteLLM gateway

[jjuritzno10]

Background

Follow-up from PR #113 review comment by @tnetennba3: #113 (comment)

The synthetic data generators under evals/synthetic/llm_generators/ (context, question, theme, response) currently instantiate openai.AsyncAzureOpenAI clients that hit Azure OpenAI directly. This deviates from Consult, where LLM requests are routed through the LiteLLM gateway.

Proposal

Point these generators at the LiteLLM gateway (potentially with a new key) so that:

• We get consistency with how Consult invokes models
• We can track and attribute spend centrally

Scope

• evals/synthetic/llm_generators/context_generator.py
• evals/synthetic/llm_generators/question_generator.py
• evals/synthetic/llm_generators/theme_generator.py
• evals/synthetic/llm_generators/response_generator.py

Notes

Not a correctness issue — synthetic data generation only — so this is intentionally a follow-up rather than a blocker for #113.