Howie/validate sample by agent

[howieleung]

The existing code has already collected print call content and validate if the content after ==> Result meets certain criteria.
Now change replace this validation by submitting all print contents to response.create and validate by AI.
This response.create call will be recorded but the input which is the print contents are sanitized. So if you modify the print statement in samples, you don't need to re-record and still able to replay the record with assertion passed.

Also, if responses said test fail, it is hard to check what content was in the print call. So I write it to temp file.

[Copilot]

Pull request overview

This PR refactors the sample testing infrastructure to use an AI agent for validating sample outputs. The main change replaces pattern-based output validation with an LLM-powered validation approach using Azure OpenAI.

Key Changes:

• Converted SampleExecutor from a simple helper class to a decorator pattern with context manager support
• Introduced agent-based validation to check if sample outputs indicate success or failure
• Moved environment variable mapping to a separate function and refactored execution flow

[dargilco]

Big change, don't forget to run 'black' tool. Thanks!