Add evaluate answer with llm-as-a-judge

[hesteban-tuenti]

This PR extends the work done in the test_text_similarity, by adding new functions more focused on querying a LLM-as-a-judge which will compare the texts (an LLM answer and a referece answer) adding to the context the user's question and optionally providing to the model a response_format object in case we need extra analysis from the LLM-as-a-judge

Changes:

• Extend openai_request to support structured parsing via response_format and normalize message construction.
• Add evaluate_answer.py with helpers to score similarity between an LLM answer and reference answer(s), plus an assertion helper.
• Add Azure OpenAI-backed pytest coverage for the new evaluation helpers (skipped unless env var is present).

Tests results:

[Copilot]

Pull request overview

Adds an “LLM-as-a-judge” answer evaluation utility on top of the existing OpenAI/Azure OpenAI integration, including optional structured output support.

Changes:

• Extend openai_request to support structured parsing via response_format and normalize message construction.
• Add evaluate_answer.py with helpers to score similarity between an LLM answer and reference answer(s), plus an assertion helper.
• Add Azure OpenAI-backed pytest coverage for the new evaluation helpers (skipped unless env var is present).

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 8 comments.

File	Description
toolium/utils/ai_utils/openai.py	Adds response_format handling and refactors message building for OpenAI/Azure calls.
toolium/utils/ai_utils/evaluate_answer.py	New module implementing LLM-as-a-judge answer evaluation + assertion helper.
toolium/test/utils/ai_utils/test_answer_evaluation.py	New integration-style tests for Azure OpenAI answer evaluation, including structured outputs.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[rgonalo]

faltaría meter algo de doc en ai_utils.rst y una línea en el CHANGELOG.rst